OpenSpace/openspace/skill_engine/evolver.py
2026-07-17 11:43:42 +08:00

1874 lines
74 KiB
Python

"""SkillEvolver authoring backend primitives.
Three evolution types:
FIX — repair broken/outdated instructions (in-place, same name)
DERIVED — create enhanced version from existing skill (new directory)
CAPTURED — capture novel reusable pattern from execution (brand new skill)
Runtime mutation must enter through TriggerJob → EvidencePacket →
DecisionRationale → Admission → staged authoring → validation → commit.
This module keeps the old authoring primitives private while staging support
is extracted.
"""
from __future__ import annotations
import asyncio
import copy
import json
import re
import shutil
import uuid
from dataclasses import dataclass, field, replace
from enum import Enum
from pathlib import Path
from typing import Any, Dict, List, Optional, Set, TYPE_CHECKING
from openspace.tool_runtime.orchestration import run_tools
from .types import (
EvolutionSuggestion,
EvolutionType,
ExecutionAnalysis,
SkillCategory,
SkillLineage,
SkillOrigin,
SkillRecord,
SkillTrustState,
)
from .patch import (
PatchType,
SkillEditResult,
collect_skill_snapshot,
create_skill,
fix_skill,
derive_skill,
SKILL_FILENAME,
)
from .skill_utils import (
get_frontmatter_field as _extract_frontmatter_field,
set_frontmatter_field as _set_frontmatter_field,
strip_markdown_fences as _strip_markdown_fences,
truncate as _truncate,
validate_skill_dir as _validate_skill_dir,
)
from .registry import write_skill_id
from .store import SkillStore
from openspace.prompts import SkillEnginePrompts
from openspace.telemetry.call_source import reset_call_source, set_call_source
from openspace.utils.logging import Logger
if TYPE_CHECKING:
from .registry import SkillRegistry
from openspace.llm import LLMClient
from openspace.grounding.core.tool import BaseTool
from openspace.grounding.core.quality.types import ToolQualityRecord
logger = Logger.get_logger(__name__)
_SKILL_CONTENT_MAX_CHARS = 12_000 # Max chars of SKILL.md in evolution prompt
_MAX_SKILL_NAME_LENGTH = 50 # Max chars for a skill name (directory name)
@dataclass
class _EvolutionFinalOutput:
edit_content: str
change_summary: Optional[str] = None
overlay_fields: Dict[str, Any] = field(default_factory=dict)
overlay_metadata: Dict[str, Dict[str, Any]] = field(default_factory=dict)
intent_spec: Dict[str, Any] = field(default_factory=dict)
eval_plan: Dict[str, Any] = field(default_factory=dict)
def _sanitize_skill_name(name: str) -> str:
"""Enforce naming rules for skill names (used as directory names).
- Lowercase, hyphens only (no underscores or special chars)
- Truncate to ``_MAX_SKILL_NAME_LENGTH`` at a word boundary
- Remove trailing hyphens
"""
# Normalize: lowercase, replace underscores and spaces with hyphens
clean = re.sub(r"[^a-z0-9\-]", "-", name.lower().strip())
# Collapse multiple hyphens
clean = re.sub(r"-{2,}", "-", clean).strip("-")
if len(clean) <= _MAX_SKILL_NAME_LENGTH:
return clean
# Truncate at a hyphen boundary to avoid cutting words
truncated = clean[:_MAX_SKILL_NAME_LENGTH]
last_hyphen = truncated.rfind("-")
if last_hyphen > _MAX_SKILL_NAME_LENGTH // 2:
truncated = truncated[:last_hyphen]
return truncated.strip("-")
def _parse_runtime_overlay_payload(
payload: Any,
) -> tuple[Dict[str, Any], Dict[str, Dict[str, Any]]]:
if not isinstance(payload, dict):
return {}, {}
raw_fields = payload.get("fields")
if not isinstance(raw_fields, dict):
raw_fields = {
key: value
for key, value in payload.items()
if key not in {"rationale", "metadata", "notes"}
}
raw_rationale = payload.get("rationale")
rationale = raw_rationale if isinstance(raw_rationale, dict) else {}
raw_metadata = payload.get("metadata")
metadata_input = raw_metadata if isinstance(raw_metadata, dict) else {}
fields: Dict[str, Any] = {}
metadata: Dict[str, Dict[str, Any]] = {}
for raw_key, value in raw_fields.items():
key = _normalize_runtime_overlay_field_name(raw_key)
if key not in _ALLOWED_RUNTIME_OVERLAY_FIELDS:
continue
normalized = _normalize_runtime_overlay_field_value(key, value)
if normalized is _INVALID_RUNTIME_OVERLAY_VALUE:
continue
fields[key] = normalized
field_meta: Dict[str, Any] = {
"risk": "high" if key in _HIGH_RISK_RUNTIME_OVERLAY_FIELDS else "low",
"source": "skill_evolver",
}
reason = rationale.get(raw_key, rationale.get(key))
if isinstance(reason, str) and reason.strip():
field_meta["rationale"] = reason.strip()
extra = metadata_input.get(raw_key, metadata_input.get(key))
if isinstance(extra, dict):
field_meta.update(extra)
metadata[key] = field_meta
return fields, metadata
def _extract_evolution_finalization(
content: str,
) -> tuple[Optional[_EvolutionFinalOutput], Optional[str], bool]:
"""Parse the structured finalization block from an evolution response.
Returns ``(output, failure_reason, found_block)``. A missing block is not a
failure by itself because the agent may still be gathering information.
"""
stripped = _strip_markdown_fences(content)
matches = list(_EVOLUTION_FINALIZATION_BLOCK_RE.finditer(stripped))
if not matches:
return None, None, False
if len(matches) != 1:
return None, "Expected exactly one evolution finalization block", True
match = matches[0]
if stripped[match.end():].strip():
return (
None,
"Evolution finalization block must be the final response content",
True,
)
raw_block = _strip_markdown_fences(match.group(1).strip())
try:
payload = json.loads(raw_block)
except Exception:
return None, "Malformed evolution finalization JSON", True
if not isinstance(payload, dict):
return None, "Evolution finalization payload must be a JSON object", True
status = str(payload.get("status", "")).strip().lower()
if status == "failed":
reason = str(payload.get("reason", "")).strip()
return None, reason or "LLM declined to produce edit", True
if status != "complete":
return None, f"Unsupported evolution finalization status: {status or '(missing)'}", True
edit_content = stripped[:match.start()].strip()
edit_content = _strip_markdown_fences(edit_content)
if not edit_content.strip():
return None, "Evolution finalization completed without edit content", True
overlay_fields: Dict[str, Any] = {}
overlay_metadata: Dict[str, Dict[str, Any]] = {}
runtime_overlay = payload.get("runtime_overlay")
if isinstance(runtime_overlay, dict):
final_fields, final_metadata = _parse_runtime_overlay_payload(runtime_overlay)
overlay_fields.update(final_fields)
overlay_metadata.update(final_metadata)
change_summary = str(payload.get("change_summary", "")).strip() or None
intent_spec = payload.get("intent_spec")
eval_plan = payload.get("eval_plan")
if not isinstance(intent_spec, dict) or not intent_spec:
return None, "Evolution finalization missing intent_spec", True
if not isinstance(eval_plan, dict) or not eval_plan:
return None, "Evolution finalization missing eval_plan", True
return (
_EvolutionFinalOutput(
edit_content=edit_content,
change_summary=change_summary,
overlay_fields=overlay_fields,
overlay_metadata=overlay_metadata,
intent_spec=dict(intent_spec),
eval_plan=dict(eval_plan),
),
None,
True,
)
def _normalize_runtime_overlay_field_name(raw_key: Any) -> str:
key = str(raw_key or "").strip()
return _RUNTIME_OVERLAY_FIELD_ALIASES.get(key, key)
_INVALID_RUNTIME_OVERLAY_VALUE = object()
def _normalize_runtime_overlay_field_value(key: str, value: Any) -> Any:
if key == "hooks":
return value if isinstance(value, dict) and value else _INVALID_RUNTIME_OVERLAY_VALUE
if key == "context":
text = str(value or "").strip().lower()
return "fork" if text == "fork" else _INVALID_RUNTIME_OVERLAY_VALUE
if key == "shell":
text = str(value or "").strip().lower()
return text if text in {"bash", "powershell"} else _INVALID_RUNTIME_OVERLAY_VALUE
if key in {"allowed-tools", "paths", "arguments"}:
if isinstance(value, (list, tuple, set)):
items = [str(item).strip() for item in value if str(item).strip()]
return items if items else _INVALID_RUNTIME_OVERLAY_VALUE
text = str(value or "").strip()
return text if text else _INVALID_RUNTIME_OVERLAY_VALUE
if key in {"disable-model-invocation", "user-invocable"}:
if isinstance(value, bool):
return value
text = str(value or "").strip().lower()
if text in {"1", "true", "yes", "y", "on"}:
return True
if text in {"0", "false", "no", "n", "off"}:
return False
return _INVALID_RUNTIME_OVERLAY_VALUE
if isinstance(value, str):
text = value.strip()
return text if text else _INVALID_RUNTIME_OVERLAY_VALUE
return value if value is not None else _INVALID_RUNTIME_OVERLAY_VALUE
_ANALYSIS_CONTEXT_MAX = 5 # Max recent analyses to include in prompt
_ANALYSIS_NOTE_MAX_CHARS = 500 # Per-analysis note truncation
# Agent loop / retry constants
_MAX_EVOLUTION_ITERATIONS = 5 # Max tool-calling rounds for evolution agent
_MAX_EVOLUTION_ATTEMPTS = 3 # Max apply-retry attempts per evolution
_MAX_EVOLUTION_LENGTH_RECOVERIES = 1
_EVOLUTION_FINALIZATION_BLOCK_RE = re.compile(
r"\n?\*\*\* Begin Evolution Finalization\s*\n(.*?)\n\*\*\* End Evolution Finalization\s*",
re.DOTALL | re.IGNORECASE,
)
def _build_evolution_length_recovery_message(
attempt: int,
) -> Dict[str, Any]:
return {
"role": "user",
"content": (
"Your previous skill-authoring response was truncated by the output "
"token limit. Do not continue it verbatim. Replace it with one "
"concise, complete response that fits: output the full SKILL.md edit "
"content, followed by exactly one Evolution Finalization block. "
"Include every required finalization field, especially status, "
"change_summary, intent_spec, and eval_plan. Keep only reusable "
"instructions and essential examples. Do not call tools. "
f"(recovery attempt {attempt}/{_MAX_EVOLUTION_LENGTH_RECOVERIES})"
),
"_meta": {
"type": "evolution_max_output_tokens_recovery",
"is_meta": True,
},
}
_LOW_RISK_RUNTIME_OVERLAY_FIELDS = {
"description",
"when_to_use",
"argument-hint",
"arguments",
"version",
"paths",
}
_HIGH_RISK_RUNTIME_OVERLAY_FIELDS = {
"allowed-tools",
"disable-model-invocation",
"user-invocable",
"model",
"effort",
"hooks",
"context",
"agent",
"shell",
}
_ALLOWED_RUNTIME_OVERLAY_FIELDS = (
_LOW_RISK_RUNTIME_OVERLAY_FIELDS | _HIGH_RISK_RUNTIME_OVERLAY_FIELDS
)
_RUNTIME_OVERLAY_FIELD_ALIASES = {
"allowed_tools": "allowed-tools",
"allowedTools": "allowed-tools",
"disable_model_invocation": "disable-model-invocation",
"disableModelInvocation": "disable-model-invocation",
"user_invocable": "user-invocable",
"userInvocable": "user-invocable",
"whenToUse": "when_to_use",
"when-to-use": "when_to_use",
"argument_hint": "argument-hint",
"argumentHint": "argument-hint",
}
class EvolutionTrigger(str, Enum):
"""What initiated this evolution."""
ANALYSIS = "analysis" # Post-execution analysis suggestion
@dataclass
class EvolutionContext:
"""Unified context for all evolution triggers.
For trigger 1 (ANALYSIS): source_task_id is set, recent_analyses may be
just the single triggering analysis.
For triggers 2/3: source_task_id is None, recent_analyses are loaded
from the skill's historical records.
"""
trigger: EvolutionTrigger
suggestion: EvolutionSuggestion
# Parent skill context
skill_records: List[SkillRecord] = field(default_factory=list)
skill_contents: List[str] = field(default_factory=list)
skill_dirs: List[Path] = field(default_factory=list)
# Task context
source_task_id: Optional[str] = None
recent_analyses: List[ExecutionAnalysis] = field(default_factory=list)
# Available tools for agent loop (read, web_search, shell, MCP, etc.)
available_tools: List["BaseTool"] = field(default_factory=list)
# For CAPTURED: preferred directory to write the new skill.
# Set from the calling host agent's skill directory so captured skills
# are written back to the correct host, not always to _skill_dirs[0].
capture_dir: Optional[Path] = None
class SkillEvolver:
"""Authoring backend primitives for skill evolution.
Runtime callers must enter through TriggerJob/EvolutionEngine. This class
keeps the legacy authoring implementation available for staged backend
extraction, but it no longer exposes public direct mutation triggers.
Concurrency:
``max_concurrent`` controls the semaphore that throttles parallel
evolutions across all trigger types. File I/O is synchronous and
naturally serialized by the event loop; only LLM calls run in
parallel.
Background:
Background task tracking remains only for in-flight tasks created by
older runtimes during shutdown; new runtime flow does not schedule
SkillEvolver direct work.
"""
def __init__(
self,
store: SkillStore,
registry: "SkillRegistry",
llm_client: "LLMClient",
model: Optional[str] = None,
available_tools: Optional[List["BaseTool"]] = None,
*,
max_tokens: Optional[int] = None,
max_concurrent: int = 3,
allow_legacy_direct_mutation: bool = False,
) -> None:
self._store = store
self._registry = registry
self._llm_client = llm_client
self._model = model
self._max_tokens = (
max(1, int(max_tokens)) if max_tokens is not None else None
)
self._available_tools: List["BaseTool"] = available_tools or []
self._allow_legacy_direct_mutation = bool(allow_legacy_direct_mutation)
# Concurrency: semaphore limits parallel LLM sessions
self._max_concurrent = max(1, max_concurrent)
self._semaphore = asyncio.Semaphore(self._max_concurrent)
# Track background tasks so they can be awaited on shutdown.
self._background_tasks: Set[asyncio.Task] = set()
def _ensure_legacy_direct_mutation_allowed(self) -> None:
if self._allow_legacy_direct_mutation:
return
raise RuntimeError(
"SkillEvolver direct mutation is disabled; use the "
"evidence-backed EvolutionEngine flow."
)
def set_available_tools(self, tools: List["BaseTool"]) -> None:
"""Update the tools available for evolution agent loops."""
self._available_tools = list(tools)
async def _call_evolution_model(
self,
*,
messages: List[Dict[str, Any]],
tools: Optional[List["BaseTool"]] = None,
model: Optional[str] = None,
):
kwargs: Dict[str, Any] = {
"messages": messages,
"tools": tools,
"model": model,
}
if self._max_tokens is not None:
kwargs["max_tokens"] = self._max_tokens
return await self._llm_client.call_model_with_fallback(**kwargs)
async def wait_background(self) -> None:
"""Await all outstanding background evolution tasks.
Call this during shutdown / cleanup to ensure nothing is lost.
"""
if self._background_tasks:
logger.info(
f"Waiting for {len(self._background_tasks)} background "
f"evolution task(s) to finish..."
)
results = await asyncio.gather(*self._background_tasks, return_exceptions=True)
for r in results:
if isinstance(r, BaseException):
logger.warning(f"Background evolution task failed during shutdown: {r}")
self._background_tasks.clear()
async def _evolve_context(self, ctx: EvolutionContext) -> Optional[SkillRecord]:
"""Execute one authoring context. Returns new SkillRecord or None.
The global semaphore is NOT acquired here — it is managed at the
caller level so the concurrency limit covers the whole batch.
"""
self._ensure_legacy_direct_mutation_allowed()
_src_tok = set_call_source("evolver")
evo_type = ctx.suggestion.evolution_type
try:
if evo_type == EvolutionType.FIX:
return await self._evolve_fix(ctx)
elif evo_type == EvolutionType.DERIVED:
return await self._evolve_derived(ctx)
elif evo_type == EvolutionType.CAPTURED:
return await self._evolve_captured(ctx)
else:
logger.warning(f"Unknown evolution type: {evo_type}")
return None
except Exception as e:
targets = "+".join(ctx.suggestion.target_skill_ids) or "(new)"
logger.error(f"Evolution failed [{evo_type.value}] target={targets}: {e}")
return None
finally:
reset_call_source(_src_tok)
async def _record_origin_trust(
self,
record: SkillRecord,
ctx: EvolutionContext,
) -> None:
recorder = getattr(self._store, "record_trust_observation", None)
if not callable(recorder):
return
task_id = str(ctx.source_task_id or "")
observation_id = (
f"task:{task_id}" if task_id else f"skill-origin:{record.skill_id}"
)
try:
observed = await recorder(
record.skill_id,
observation_id,
"success",
task_id=task_id,
source="legacy_evolution_origin",
)
if isinstance(observed, SkillRecord):
record.trust_state = observed.trust_state
record.trust_successes = observed.trust_successes
record.trust_failures = observed.trust_failures
record.last_updated = observed.last_updated
except Exception:
logger.warning(
"Evolution trust origin record failed for %s",
record.skill_id,
exc_info=True,
)
async def _suggest_runtime_overlays_from_analysis(
self,
analysis: ExecutionAnalysis,
) -> None:
"""Write conservative runtime-field overlay suggestions from outcomes.
OpenSpace's evolve loop can improve skill manifests without mutating
community/plugin skill files. This method only suggests low-risk
fields and never auto-approves permission, hook, shell, model, effort,
context, or agent changes.
"""
if not analysis.skill_judgments:
return
phase_failed = set(analysis.skill_phase_failed_skill_ids or [])
note = " ".join(str(analysis.execution_note or "").split())
for judgment in analysis.skill_judgments:
if not judgment.skill_applied:
continue
if not analysis.task_completed or judgment.skill_id in phase_failed:
continue
try:
meta = self._registry.get_skill(judgment.skill_id)
except Exception:
meta = None
if meta is None:
continue
suggestions: dict[str, Any] = {}
if not getattr(meta, "when_to_use", None):
basis = note or judgment.note or f"successful task {analysis.task_id}"
suggestions["when_to_use"] = _truncate(
f"Use when this workflow matches: {basis}",
240,
)
if not suggestions:
continue
try:
self._registry.write_runtime_overlay(
judgment.skill_id,
suggestions,
approved=False,
field_metadata={
key: {
"risk": "low",
"source": "skill_evolver.analysis",
"rationale": (
"Successful applied skill lacked low-risk "
"runtime selection metadata."
),
}
for key in suggestions
},
)
except Exception:
logger.debug(
"Runtime overlay suggestion skipped for %s",
judgment.skill_id,
exc_info=True,
)
continue
try:
await self._store.record_skill_event(
judgment.skill_id,
"field_suggested",
source="evolver_overlay",
task_id=analysis.task_id,
metadata={
"fields": sorted(suggestions),
"approved": False,
"reason": "successful applied skill lacked low-risk runtime metadata",
},
)
except Exception:
logger.debug(
"Runtime overlay field_suggested event skipped for %s",
judgment.skill_id,
exc_info=True,
)
async def _write_generated_runtime_overlay(
self,
skill_id: str,
fields: Dict[str, Any],
metadata: Dict[str, Dict[str, Any]],
ctx: EvolutionContext,
) -> None:
"""Persist LLM-generated runtime overlay suggestions for human review."""
if not fields:
return
enriched_meta: Dict[str, Dict[str, Any]] = {}
for key in fields:
base = dict(metadata.get(key) or {})
base.setdefault(
"risk",
"high" if key in _HIGH_RISK_RUNTIME_OVERLAY_FIELDS else "low",
)
base.setdefault("source", "skill_evolver.generated")
base["evolution_type"] = ctx.suggestion.evolution_type.value
if ctx.source_task_id:
base["source_task_id"] = ctx.source_task_id
enriched_meta[key] = base
try:
self._registry.write_runtime_overlay(
skill_id,
fields,
approved=False,
field_metadata=enriched_meta,
)
except Exception:
logger.debug(
"Generated runtime overlay suggestion skipped for %s",
skill_id,
exc_info=True,
)
return
try:
await self._store.record_skill_event(
skill_id,
"field_suggested",
source="skill_evolver_generated_overlay",
task_id=ctx.source_task_id or "",
metadata={
"fields": sorted(fields),
"approved": False,
"high_risk_fields": sorted(
key for key in fields if key in _HIGH_RISK_RUNTIME_OVERLAY_FIELDS
),
"reason": "LLM emitted structured runtime overlay suggestions",
},
)
except Exception:
logger.debug(
"Generated runtime overlay field_suggested event skipped for %s",
skill_id,
exc_info=True,
)
async def _execute_contexts(
self,
contexts: List[EvolutionContext],
trigger_label: str,
) -> List[SkillRecord]:
"""Execute a list of evolution contexts in parallel (throttled).
Used by all three triggers after building/confirming contexts.
"""
self._ensure_legacy_direct_mutation_allowed()
async def _throttled(c: EvolutionContext) -> Optional[SkillRecord]:
async with self._semaphore:
return await self._evolve_context(c)
raw = await asyncio.gather(
*[_throttled(c) for c in contexts],
return_exceptions=True,
)
results: List[SkillRecord] = []
for r in raw:
if isinstance(r, BaseException):
logger.error(f"[Trigger:{trigger_label}] Evolution task raised: {r}")
elif r is not None:
results.append(r)
if results:
names = [r.name for r in results]
logger.info(
f"[Trigger:{trigger_label}] Evolved {len(results)} skill(s): {names}"
)
return results
def schedule_background(
self,
coro,
*,
label: str = "background_evolution",
) -> Optional[asyncio.Task]:
"""Launch a coroutine as a background ``asyncio.Task``.
Used by the runtime post-execution quality evolution path when
``background_triggers`` is True. The task is tracked so it can
be awaited on shutdown via ``wait_background()``.
"""
try:
loop = asyncio.get_running_loop()
except RuntimeError:
logger.warning(f"No running event loop — cannot schedule {label}")
return None
task = loop.create_task(coro, name=label)
self._background_tasks.add(task)
task.add_done_callback(self._background_tasks.discard)
task.add_done_callback(self._log_background_result)
return task
@staticmethod
def _log_background_result(task: asyncio.Task) -> None:
"""Log the outcome of a background evolution task."""
if task.cancelled():
logger.debug(f"Background task '{task.get_name()}' was cancelled")
return
exc = task.exception()
if exc:
logger.error(
f"Background task '{task.get_name()}' failed: {exc}",
exc_info=exc,
)
async def _evolve_fix(self, ctx: EvolutionContext) -> Optional[SkillRecord]:
"""In-place fix: same name, same directory, new version record.
Uses explicit ``call_model()`` + ``run_tools()`` turns for
information gathering before the apply-retry cycle.
"""
self._ensure_legacy_direct_mutation_allowed()
if not ctx.skill_records or not ctx.skill_contents or not ctx.skill_dirs:
logger.warning("FIX requires exactly 1 parent (skill_records/contents/dirs)")
return None
parent = ctx.skill_records[0]
parent_content = ctx.skill_contents[0]
parent_dir = ctx.skill_dirs[0]
# Build prompt with full directory content for multi-file skills
dir_content = self._format_skill_dir_content(parent_dir)
prompt = SkillEnginePrompts.evolution_fix(
current_content=_truncate(dir_content or parent_content, _SKILL_CONTENT_MAX_CHARS),
direction=ctx.suggestion.direction,
failure_context=self._format_analysis_context(ctx.recent_analyses),
)
# Agent loop: LLM can gather information via tools before generating edits
evolution_output = await self._run_evolution_loop(prompt, ctx)
if not evolution_output:
return None
new_content = evolution_output.edit_content
overlay_fields = evolution_output.overlay_fields
overlay_metadata = evolution_output.overlay_metadata
change_summary = evolution_output.change_summary
# Apply-retry cycle
edit_result = await self._apply_with_retry(
apply_fn=lambda content: fix_skill(parent_dir, content, PatchType.AUTO),
initial_content=new_content,
skill_dir=parent_dir,
ctx=ctx,
prompt=prompt,
)
if edit_result is None or not edit_result.ok:
return None
# Re-read name/description from the updated SKILL.md on disk —
# the LLM may have refined the description (or even name) during the fix.
updated_skill_md = edit_result.content_snapshot.get(SKILL_FILENAME, "")
fixed_name = _extract_frontmatter_field(updated_skill_md, "name") or parent.name
fixed_desc = _extract_frontmatter_field(updated_skill_md, "description") or parent.description
new_id = f"{fixed_name}__v{parent.lineage.generation + 1}_{uuid.uuid4().hex[:8]}"
model = self._model or self._llm_client.model
new_record = SkillRecord(
skill_id=new_id,
name=fixed_name,
description=fixed_desc,
path=parent.path,
trust_state=SkillTrustState.PROVISIONAL,
category=parent.category,
tags=list(parent.tags),
visibility=parent.visibility,
creator_id=parent.creator_id,
lineage=SkillLineage(
origin=SkillOrigin.FIXED,
generation=parent.lineage.generation + 1,
parent_skill_ids=[parent.skill_id],
source_task_id=ctx.source_task_id,
change_summary=change_summary or ctx.suggestion.direction,
content_diff=edit_result.content_diff,
content_snapshot=edit_result.content_snapshot,
created_by=model,
),
tool_dependencies=list(parent.tool_dependencies),
critical_tools=list(parent.critical_tools),
)
await self._store.evolve_skill(new_record, [parent.skill_id])
# Stamp the new skill_id into the sidecar file so next discover()
write_skill_id(parent_dir, new_id)
await self._write_generated_runtime_overlay(
new_id,
overlay_fields,
overlay_metadata,
ctx,
)
new_meta = self._registry.load_skill_from_dir(parent_dir)
if new_meta is not None:
self._registry.update_skill(parent.skill_id, new_meta)
logger.info(
f"FIX: {parent.name} gen{parent.lineage.generation}"
f"gen{new_record.lineage.generation} [{new_id}]"
)
return new_record
async def _evolve_derived(self, ctx: EvolutionContext) -> Optional[SkillRecord]:
"""Create enhanced version in a new directory.
Supports single-parent (enhance) and multi-parent (merge/fuse).
Uses agent loop for information gathering + apply-retry cycle.
"""
self._ensure_legacy_direct_mutation_allowed()
if not ctx.skill_records or not ctx.skill_contents or not ctx.skill_dirs:
logger.warning("DERIVED requires at least one parent skill_record + content + dir")
return None
first_parent = ctx.skill_records[0] # For fallback defaults only
is_merge = len(ctx.skill_records) > 1
# Build prompt — include all parent contents for multi-parent merge
if is_merge:
parent_sections = []
for i, (rec, sd) in enumerate(zip(ctx.skill_records, ctx.skill_dirs)):
dir_content = self._format_skill_dir_content(sd)
label = f"Parent {i + 1}: {rec.name}"
parent_sections.append(
f"## {label}\n{_truncate(dir_content or ctx.skill_contents[i], _SKILL_CONTENT_MAX_CHARS)}"
)
combined_content = "\n\n---\n\n".join(parent_sections)
else:
dir_content = self._format_skill_dir_content(ctx.skill_dirs[0])
combined_content = _truncate(dir_content or ctx.skill_contents[0], _SKILL_CONTENT_MAX_CHARS)
prompt = SkillEnginePrompts.evolution_derived(
parent_content=combined_content,
direction=ctx.suggestion.direction,
execution_insights=self._format_analysis_context(ctx.recent_analyses),
)
# Agent loop
evolution_output = await self._run_evolution_loop(prompt, ctx)
if not evolution_output:
return None
new_content = evolution_output.edit_content
overlay_fields = evolution_output.overlay_fields
overlay_metadata = evolution_output.overlay_metadata
change_summary = evolution_output.change_summary
# Determine new skill name from frontmatter, or generate one
new_name = _extract_frontmatter_field(new_content, "name")
if not new_name or new_name == first_parent.name:
suffix = "-merged" if is_merge else "-enhanced"
new_name = f"{first_parent.name}{suffix}"
new_content = _set_frontmatter_field(new_content, "name", new_name)
# Cap name length to avoid ever-growing chains like
# "panel-component-enhanced-enhanced-merged_abc123"
new_name = _sanitize_skill_name(new_name)
new_content = _set_frontmatter_field(new_content, "name", new_name)
# Directory name always matches the skill name
target_dir = ctx.skill_dirs[0].parent / new_name
if target_dir.exists():
new_name = f"{new_name}-{uuid.uuid4().hex[:6]}"
new_name = _sanitize_skill_name(new_name)
target_dir = ctx.skill_dirs[0].parent / new_name
new_content = _set_frontmatter_field(new_content, "name", new_name)
# Apply-retry cycle for derive_skill
edit_result = await self._apply_with_retry(
apply_fn=lambda content: derive_skill(ctx.skill_dirs, target_dir, content, PatchType.AUTO),
initial_content=new_content,
skill_dir=target_dir,
ctx=ctx,
prompt=prompt,
cleanup_on_retry=target_dir, # Remove failed target dir before retry
)
if edit_result is None or not edit_result.ok:
return None
# Extract description from new content
new_desc = _extract_frontmatter_field(new_content, "description") or first_parent.description
# Collect parent info from ALL parents
parent_ids = [r.skill_id for r in ctx.skill_records]
max_gen = max(r.lineage.generation for r in ctx.skill_records)
all_tool_deps: set = set()
all_critical: set = set()
all_tags: set = set()
for rec in ctx.skill_records:
all_tool_deps.update(rec.tool_dependencies)
all_critical.update(rec.critical_tools)
all_tags.update(rec.tags)
new_id = f"{new_name}__v0_{uuid.uuid4().hex[:8]}"
model = self._model or self._llm_client.model
new_record = SkillRecord(
skill_id=new_id,
name=new_name,
description=new_desc,
path=str(target_dir / SKILL_FILENAME),
trust_state=SkillTrustState.PROVISIONAL,
category=ctx.suggestion.category or first_parent.category,
tags=sorted(all_tags),
visibility=first_parent.visibility,
creator_id=first_parent.creator_id,
lineage=SkillLineage(
origin=SkillOrigin.DERIVED,
generation=max_gen + 1,
parent_skill_ids=parent_ids,
source_task_id=ctx.source_task_id,
change_summary=change_summary or ctx.suggestion.direction,
content_diff=edit_result.content_diff,
content_snapshot=edit_result.content_snapshot,
created_by=model,
),
tool_dependencies=sorted(all_tool_deps),
critical_tools=sorted(all_critical),
)
await self._store.evolve_skill(new_record, parent_ids)
await self._record_origin_trust(new_record, ctx)
# Stamp skill_id sidecar so discover() uses this ID on restart
write_skill_id(target_dir, new_id)
target_dir = await self._materialize_generated_local_taxonomy(
record=new_record,
target_dir=target_dir,
ctx=ctx,
parent_skill_ids=parent_ids,
action_type="DERIVED",
)
await self._write_generated_runtime_overlay(
new_id,
overlay_fields,
overlay_metadata,
ctx,
)
# Register the new skill so it's immediately available with the full
# OpenSpace runtime contract parsed from its SKILL.md frontmatter.
new_meta = self._registry.load_skill_from_dir(target_dir)
if new_meta is not None:
self._registry.add_skill(new_meta)
parent_names = " + ".join(r.name for r in ctx.skill_records)
logger.info(f"DERIVED: {parent_names}{new_name} [{new_id}]")
return new_record
async def _evolve_captured(self, ctx: EvolutionContext) -> Optional[SkillRecord]:
"""Capture a novel pattern as a brand-new skill.
Uses agent loop for information gathering + apply-retry cycle.
"""
self._ensure_legacy_direct_mutation_allowed()
# Build prompt and call LLM
# For CAPTURED, we use analyses as context (the tasks where the pattern was observed)
task_descriptions = []
for a in ctx.recent_analyses[:_ANALYSIS_CONTEXT_MAX]:
if a.execution_note:
task_descriptions.append(
f"- task={a.task_id}: {a.execution_note[:200]}"
)
prompt = SkillEnginePrompts.evolution_captured(
direction=ctx.suggestion.direction,
category=(ctx.suggestion.category or SkillCategory.WORKFLOW).value,
execution_highlights="\n".join(task_descriptions) if task_descriptions else "(no task context available)",
)
# Agent loop
evolution_output = await self._run_evolution_loop(prompt, ctx)
if not evolution_output:
return None
new_content = evolution_output.edit_content
overlay_fields = evolution_output.overlay_fields
overlay_metadata = evolution_output.overlay_metadata
change_summary = evolution_output.change_summary
# Extract name/description from the generated content
new_name = _extract_frontmatter_field(new_content, "name")
new_desc = _extract_frontmatter_field(new_content, "description")
if not new_name:
logger.warning("CAPTURED: LLM did not produce a valid skill name")
return None
# Sanitize name (enforce length limit + valid chars)
new_name = _sanitize_skill_name(new_name)
new_content = _set_frontmatter_field(new_content, "name", new_name)
# Create new skill directory via create_skill (handles multi-file FULL)
# Priority chain for choosing the target skill root:
# 1. ctx.capture_dir — explicitly set from host agent's skill_dirs param
# 2. Infer from analysis — if this task used a skill from dir B,
# captured skills belong alongside it (same host agent context)
# 3. registry._skill_dirs[0] — ultimate fallback
base_dir: Optional[Path] = None
if ctx.capture_dir and ctx.capture_dir.is_dir():
base_dir = ctx.capture_dir
else:
base_dir = self._infer_capture_dir_from_analysis(ctx)
if base_dir is None:
skill_dirs = self._registry._skill_dirs
if not skill_dirs:
logger.warning("CAPTURED: no skill directories configured")
return None
base_dir = skill_dirs[0]
target_dir = base_dir / new_name
if target_dir.exists():
new_name = f"{new_name}-{uuid.uuid4().hex[:6]}"
new_name = _sanitize_skill_name(new_name)
target_dir = base_dir / new_name
new_content = _set_frontmatter_field(new_content, "name", new_name)
# Apply-retry cycle for create_skill
edit_result = await self._apply_with_retry(
apply_fn=lambda content: create_skill(target_dir, content, PatchType.AUTO),
initial_content=new_content,
skill_dir=target_dir,
ctx=ctx,
prompt=prompt,
cleanup_on_retry=target_dir,
)
if edit_result is None or not edit_result.ok:
return None
snapshot = edit_result.content_snapshot
add_all_diff = edit_result.content_diff
new_id = f"{new_name}__v0_{uuid.uuid4().hex[:8]}"
model = self._model or self._llm_client.model
new_record = SkillRecord(
skill_id=new_id,
name=new_name,
description=new_desc or new_name,
path=str(target_dir / SKILL_FILENAME),
trust_state=SkillTrustState.PROVISIONAL,
category=ctx.suggestion.category or SkillCategory.WORKFLOW,
lineage=SkillLineage(
origin=SkillOrigin.CAPTURED,
generation=0,
parent_skill_ids=[],
source_task_id=ctx.source_task_id,
change_summary=change_summary or ctx.suggestion.direction,
content_diff=add_all_diff,
content_snapshot=snapshot,
created_by=model,
),
)
await self._store.save_record(new_record)
await self._record_origin_trust(new_record, ctx)
# Stamp skill_id sidecar so discover() uses this ID on restart
write_skill_id(target_dir, new_id)
target_dir = await self._materialize_generated_local_taxonomy(
record=new_record,
target_dir=target_dir,
ctx=ctx,
parent_skill_ids=[],
action_type="CAPTURED",
)
await self._write_generated_runtime_overlay(
new_id,
overlay_fields,
overlay_metadata,
ctx,
)
# Register the new skill so it's immediately available with the full
# OpenSpace runtime contract parsed from its SKILL.md frontmatter.
new_meta = self._registry.load_skill_from_dir(target_dir)
if new_meta is not None:
self._registry.add_skill(new_meta)
logger.info(f"CAPTURED: {new_name} [{new_id}]")
return new_record
async def _materialize_generated_local_taxonomy(
self,
*,
record: SkillRecord,
target_dir: Path,
ctx: EvolutionContext,
parent_skill_ids: list[str],
action_type: str,
) -> Path:
try:
from openspace.cloud.local_mapping import CloudLocalMappingStore
from openspace.cloud.skill_classification import (
build_local_category_path,
classify_skill_dir,
initialize_local_skill_taxonomy,
materialize_skill_category_tree,
persist_skill_classification,
)
db_path = getattr(self._store, "db_path", None)
if db_path is None and getattr(self._store, "base", None) is not None:
db_path = getattr(self._store.base, "db_path", None)
mapping_store = CloudLocalMappingStore(db_path)
try:
parent_records = []
for parent_id in parent_skill_ids:
parent_record = self._store.load_record(parent_id)
if parent_record is not None:
parent_records.append(parent_record)
if parent_records:
initialize_local_skill_taxonomy(
mapping_store=mapping_store,
skills=parent_records,
)
parent_classification = None
parent_cloud_path = ""
for parent_id in parent_skill_ids:
parent_classification = mapping_store.get_skill_local_classification(parent_id)
parent_binding = mapping_store.get_skill_cloud_binding_by_local(parent_id)
if parent_binding is not None and not parent_cloud_path:
parent_cloud_path = (
parent_binding.current_package_path
or parent_binding.package_path_at_pull
or ""
)
if parent_classification is not None:
break
decision_path = str(getattr(ctx.suggestion, "local_category_path", "") or "").strip()
inherited_path = (
parent_classification.local_category_path
if parent_classification is not None
and parent_classification.local_category_path
and action_type == "DERIVED"
else ""
)
selected_path = decision_path or inherited_path
category = record.category.value
local_category_path = build_local_category_path(
category,
local_category_path=selected_path,
cloud_package_path=parent_cloud_path or None,
local_path=str(target_dir),
name=record.name,
)
classification = classify_skill_dir(
target_dir,
local_skill_id=record.skill_id,
cloud_package_path=parent_cloud_path or None,
local_category=category,
local_category_path=local_category_path,
origin="derive" if action_type == "DERIVED" else "capture",
)
classification = replace(
classification,
category=category,
local_category_path=local_category_path,
evidence={
**dict(classification.evidence or {}),
"origin": "derive" if action_type == "DERIVED" else "capture",
"evolution_action_type": action_type,
"parent_skill_ids": list(parent_skill_ids),
},
)
saved = persist_skill_classification(mapping_store, classification)
materialized_dir = materialize_skill_category_tree(
target_dir,
saved,
skills_root=target_dir.parent,
)
if materialized_dir != target_dir:
record.path = str(materialized_dir / SKILL_FILENAME)
await self._store.save_record(record)
return materialized_dir
finally:
mapping_store.close()
except Exception as exc:
logger.debug("generated local taxonomy materialization skipped: %s", exc)
return target_dir
def _infer_capture_dir_from_analysis(
self, ctx: EvolutionContext,
) -> Optional[Path]:
"""Infer the best skill root for a CAPTURED skill from analysis context.
When ``capture_dir`` is not explicitly set (no ``skill_dirs`` param
from the host agent), we look at which skills were used during the
task that triggered the capture. If a used skill lives under one
of the registered skill roots, that root is a reasonable home for
the new captured skill (same host agent context).
"""
if not ctx.recent_analyses:
return None
registry_roots = self._registry._skill_dirs
if not registry_roots:
return None
for analysis in ctx.recent_analyses:
for judgment in analysis.skill_judgments:
if not judgment.skill_applied:
continue
rec = self._store.load_record(judgment.skill_id)
if not rec or not rec.path:
continue
skill_path = Path(rec.path).parent # e.g. /A/foo/
for root in registry_roots:
try:
skill_path.relative_to(root)
logger.debug(
"CAPTURED: inferred capture dir %s from "
"applied skill %s", root, judgment.skill_id,
)
return root
except ValueError:
continue
return None
async def _run_evolution_loop(
self,
prompt: str,
ctx: EvolutionContext,
) -> Optional[_EvolutionFinalOutput]:
"""Run evolution as a structured-finalization agent loop.
Modeled after ``GroundingAgent.process()`` — the loop continues
until the LLM outputs an explicit finalization JSON block, NOT
based on whether tools were called.
Termination signals (checked every iteration, regardless of tool use):
- finalization ``status=complete`` → success, return edit payload.
- finalization ``status=failed`` → failure, return None.
Tool availability:
- Iterations 1 … N-1: tools enabled (LLM may gather information).
- Iteration N (final): tools disabled, LLM must output a decision.
Each non-final iteration without finalization gets a nudge message
telling the LLM which iteration it is on and how many remain.
Conversations are recorded to ``conversations.jsonl`` via
``RecordingManager`` (agent_name="SkillEvolver") so the full
evolution dialogue is preserved for debugging and replay.
"""
from openspace.recording import RecordingManager
model = self._model or self._llm_client.model
# Merge tools from context and instance-level
evolution_tools: List["BaseTool"] = list(ctx.available_tools or [])
if not evolution_tools:
evolution_tools = list(self._available_tools)
messages: List[Dict[str, Any]] = [
{"role": "user", "content": prompt},
]
# Record initial conversation setup
await RecordingManager.record_conversation_setup(
setup_messages=copy.deepcopy(messages),
tools=evolution_tools if evolution_tools else None,
agent_name="SkillEvolver",
extra={
"evolution_type": ctx.suggestion.evolution_type.value,
"trigger": ctx.trigger.value,
"target_skills": ctx.suggestion.target_skill_ids,
},
)
length_recovery_count = 0
for iteration in range(_MAX_EVOLUTION_ITERATIONS):
is_last = iteration == _MAX_EVOLUTION_ITERATIONS - 1
is_length_recovery = length_recovery_count > 0
# Snapshot message count before any additions + LLM call
msg_count_before = len(messages)
# Final round: disable tools and force a decision
if is_last:
messages.append({
"role": "system",
"content": (
f"This is your FINAL round (iteration "
f"{iteration + 1}/{_MAX_EVOLUTION_ITERATIONS}) — "
f"no more tool calls allowed. "
f"You MUST output either the skill edit content followed "
f"by a structured Evolution Finalization block with "
f'\"status\": \"complete\", or only a finalization block '
f'with \"status\": \"failed\" and a reason. Follow the '
f"output format specified in the original instructions."
),
})
try:
model_response = await self._call_evolution_model(
messages=messages,
tools=(
evolution_tools
if (evolution_tools and not is_last and not is_length_recovery)
else None
),
model=model,
)
except Exception as e:
logger.error(f"Evolution LLM call failed (iter {iteration + 1}): {e}")
return None
# Keep fallback state local to this evolution run. The helper
# exposes the model that produced the response without mutating
# the shared client defaults.
model = model_response.effective_model or model
assistant_message = model_response.assistant_message
messages.append(assistant_message)
raw_content = assistant_message.get("content", "")
content = (
raw_content if isinstance(raw_content, str) else str(raw_content or "")
)
has_tool_calls = bool(model_response.tool_calls)
followup_messages = self._llm_client.get_model_response_followup_messages(
model_response
)
has_api_error = self._llm_client.model_response_has_api_error(
model_response
)
recover_length = bool(
has_api_error
and model_response.stop_reason == "length"
and not has_tool_calls
and not is_last
and length_recovery_count < _MAX_EVOLUTION_LENGTH_RECOVERIES
)
if recover_length:
length_recovery_count += 1
messages.append(
_build_evolution_length_recovery_message(length_recovery_count)
)
elif followup_messages:
messages.extend(followup_messages)
tool_results: list[dict[str, Any]] = []
prevent_continuation = False
tool_stop_reason: str | None = None
if not has_api_error and has_tool_calls and evolution_tools and not is_last:
tool_context = self._llm_client.build_auxiliary_tool_use_context(
tools=evolution_tools,
messages=messages,
model=model,
agent_id="skill_evolver",
agent_type=self.__class__.__name__,
task_description=prompt,
current_iteration=iteration + 1,
max_iterations=_MAX_EVOLUTION_ITERATIONS,
)
tools_result = await run_tools(
model_response.tool_calls,
model_response.tool_map,
tool_context,
assistant_message=assistant_message,
)
messages.extend(tools_result.messages)
prevent_continuation = tools_result.prevent_continuation
tool_stop_reason = tools_result.stop_reason
tool_results = self._llm_client.collect_tool_results(
model_response.tool_calls,
model_response.tool_map,
tools_result.messages,
)
updated_messages = messages
# Record iteration delta
delta = updated_messages[msg_count_before:]
await RecordingManager.record_iteration_context(
iteration=iteration + 1,
delta_messages=copy.deepcopy(delta),
response_metadata={
"has_tool_calls": has_tool_calls,
"tool_calls_count": len(tool_results),
"has_finalization_block": bool(
content and _EVOLUTION_FINALIZATION_BLOCK_RE.search(content)
),
"prevent_continuation": prevent_continuation,
"stop_reason": model_response.stop_reason,
"tool_stop_reason": tool_stop_reason,
"content_length": len(content),
"output_tokens": int(
getattr(model_response.usage, "output_tokens", 0) or 0
),
"effective_model": model_response.effective_model or model,
"has_api_error": has_api_error,
"api_error_followup_count": len(followup_messages),
"length_recovery_attempt": (
length_recovery_count if recover_length else 0
),
},
agent_name="SkillEvolver",
)
messages = updated_messages
if has_api_error:
if recover_length:
logger.warning(
"Evolution LLM response was truncated; retrying with "
"compact finalization (iter=%s stop_reason=%s "
"output_tokens=%s content_chars=%s recovery=%s/%s)",
iteration + 1,
model_response.stop_reason or "unknown",
int(getattr(model_response.usage, "output_tokens", 0) or 0),
len(content),
length_recovery_count,
_MAX_EVOLUTION_LENGTH_RECOVERIES,
)
continue
logger.warning(
"Evolution LLM returned a non-recoverable API error "
"followup (iter=%s stop_reason=%s output_tokens=%s "
"content_chars=%s)",
iteration + 1,
model_response.stop_reason or "unknown",
int(getattr(model_response.usage, "output_tokens", 0) or 0),
len(content),
)
return None
if prevent_continuation:
logger.warning(
"Evolution agent stopped after tool hooks prevented "
"continuation: %s",
tool_stop_reason or "hook_stopped",
)
return None
# ── Finalization check (every iteration, regardless of tool calls) ──
if content:
final_output, failure_reason, found_finalization = (
_extract_evolution_finalization(content)
)
if failure_reason is not None:
targets = "+".join(ctx.suggestion.target_skill_ids) or "(new)"
logger.warning(
f"Evolution LLM signalled failure "
f"[{ctx.suggestion.evolution_type.value}] "
f"target={targets}: {failure_reason}"
)
return None
if final_output is not None:
return final_output
if found_finalization:
return None
# No finalization found
if is_last:
# Final round exhausted without a decision
logger.warning(
f"Evolution agent finished {_MAX_EVOLUTION_ITERATIONS} iterations "
f"without a structured finalization block"
)
return None
if has_tool_calls:
logger.debug(
f"Evolution agent used tools "
f"(iter {iteration + 1}/{_MAX_EVOLUTION_ITERATIONS})"
)
else:
# No tools, no finalization — nudge the LLM
logger.debug(
f"Evolution agent produced content without finalization or tools "
f"(iter {iteration + 1}/{_MAX_EVOLUTION_ITERATIONS})"
)
# Iteration guidance
remaining = _MAX_EVOLUTION_ITERATIONS - iteration - 1
messages.append({
"role": "system",
"content": (
f"Iteration {iteration + 1}/{_MAX_EVOLUTION_ITERATIONS} complete "
f"({remaining} remaining). "
f"If your edit is ready, output it followed by an Evolution "
f"Finalization block with status=complete. "
f"If you cannot complete this evolution, output only an "
f"Evolution Finalization block with status=failed and a reason. "
f"Otherwise, continue gathering information with tools."
),
})
# Should never reach here (is_last handles the final iteration)
return None
async def _apply_with_retry(
self,
*,
apply_fn,
initial_content: str,
skill_dir: Path,
ctx: EvolutionContext,
prompt: str,
cleanup_on_retry: Optional[Path] = None,
) -> Optional[SkillEditResult]:
"""Apply an edit with retry on failure.
If the first attempt fails (patch parse error, path mismatch, etc.),
feeds the error back to the LLM and asks for a corrected version.
After successful application, runs structural validation.
Retry conversations are recorded to ``conversations.jsonl`` under
agent_name="SkillEvolver.retry" so failed apply attempts and LLM
corrections are preserved for debugging.
Args:
apply_fn: Callable that takes content str and returns SkillEditResult.
initial_content: First LLM-generated content to try.
skill_dir: Skill directory for validation.
ctx: Evolution context (for retry LLM calls).
prompt: Original prompt (for retry context).
cleanup_on_retry: Directory to remove before retrying (for derive/create).
"""
from openspace.recording import RecordingManager
current_content = initial_content
msg_history: List[Dict[str, Any]] = [
{"role": "user", "content": prompt},
{"role": "assistant", "content": initial_content},
]
# Track whether we've recorded the retry setup (only on first retry)
retry_setup_recorded = False
model = self._model or self._llm_client.model
for attempt in range(_MAX_EVOLUTION_ATTEMPTS):
# Clean up previous failed attempt (for derive/create)
if attempt > 0 and cleanup_on_retry and cleanup_on_retry.exists():
shutil.rmtree(cleanup_on_retry, ignore_errors=True)
# Apply the edit
edit_result = apply_fn(current_content)
if edit_result.ok:
# Validate the result
validation_error = _validate_skill_dir(skill_dir)
if validation_error is None:
if attempt > 0:
logger.info(
f"Apply-retry succeeded on attempt {attempt + 1}/{_MAX_EVOLUTION_ATTEMPTS}"
)
return edit_result
else:
# Validation failed — treat as error for retry
error_msg = f"Validation failed: {validation_error}"
logger.warning(
f"Apply succeeded but validation failed "
f"(attempt {attempt + 1}/{_MAX_EVOLUTION_ATTEMPTS}): "
f"{validation_error}"
)
else:
error_msg = edit_result.error or "Unknown apply error"
logger.warning(
f"Apply failed (attempt {attempt + 1}/{_MAX_EVOLUTION_ATTEMPTS}): "
f"{error_msg}"
)
# Last attempt? Give up.
if attempt >= _MAX_EVOLUTION_ATTEMPTS - 1:
logger.error(
f"Apply-retry exhausted after {_MAX_EVOLUTION_ATTEMPTS} attempts. "
f"Last error: {error_msg}"
)
# Clean up any partially created directory
if cleanup_on_retry and cleanup_on_retry.exists():
shutil.rmtree(cleanup_on_retry, ignore_errors=True)
return None
# Record retry setup on first retry attempt
if not retry_setup_recorded:
await RecordingManager.record_conversation_setup(
setup_messages=copy.deepcopy(msg_history),
agent_name="SkillEvolver.retry",
extra={
"evolution_type": ctx.suggestion.evolution_type.value,
"target_skills": ctx.suggestion.target_skill_ids,
"first_error": error_msg[:300],
},
)
retry_setup_recorded = True
# Feed error back to LLM for retry, including current file
# content so the LLM doesn't hallucinate what's on disk.
current_on_disk = self._format_skill_dir_content(skill_dir) if skill_dir.is_dir() else ""
retry_prompt = (
f"The previous edit was not successful. "
f"This was the error:\n\n{error_msg}\n\n"
)
if current_on_disk:
retry_prompt += (
f"Here is the CURRENT content of the skill files on disk "
f"(use this as the ground truth for any SEARCH/REPLACE or "
f"context anchors):\n\n{_truncate(current_on_disk, _SKILL_CONTENT_MAX_CHARS)}\n\n"
)
retry_prompt += (
f"Please fix the issue and generate the edit again. "
f"Follow the same output format as before."
)
msg_history.append({"role": "user", "content": retry_prompt})
# Call LLM for corrected version (no tools — just fix the edit)
try:
result = await self._call_evolution_model(
messages=msg_history,
model=model,
)
model = result.effective_model or model
new_content = result.assistant_message.get("content", "")
if not new_content:
logger.warning("Retry LLM returned empty content")
continue
final_output, failure_reason, found_finalization = (
_extract_evolution_finalization(new_content)
)
if failure_reason is not None:
logger.warning(
"Retry LLM signalled evolution failure: %s",
failure_reason,
)
return None
if final_output is None:
if found_finalization:
logger.warning("Retry LLM returned unusable finalization")
else:
logger.warning("Retry LLM did not return finalization block")
continue
new_content = final_output.edit_content
msg_history.append({"role": "assistant", "content": new_content})
current_content = new_content
# Record retry iteration
await RecordingManager.record_iteration_context(
iteration=attempt + 1,
delta_messages=[
{"role": "user", "content": retry_prompt},
{"role": "assistant", "content": new_content},
],
response_metadata={
"has_tool_calls": False,
"attempt": attempt + 1,
"error": error_msg[:300],
},
agent_name="SkillEvolver.retry",
)
except Exception as e:
logger.error(f"Retry LLM call failed: {e}")
continue
return None
def _build_context_from_analysis(
self,
analysis: ExecutionAnalysis,
suggestion: EvolutionSuggestion,
capture_dir: Optional[Path] = None,
) -> Optional[EvolutionContext]:
"""Build EvolutionContext from a single analysis suggestion.
Loads all target skills referenced by ``suggestion.target_skill_ids``.
For FIX: exactly 1 parent required.
For DERIVED: 1+ parents (multi-parent = merge).
For CAPTURED: parents list is empty; ``capture_dir`` controls where
the new skill is written (defaults to registry's first skill root).
"""
records: List[SkillRecord] = []
contents: List[str] = []
dirs: List[Path] = []
if suggestion.evolution_type in (EvolutionType.FIX, EvolutionType.DERIVED):
if not suggestion.target_skill_ids:
logger.warning("FIX/DERIVED suggestion missing target_skill_ids")
return None
resolved_target_ids: List[str] = []
for target_id in suggestion.target_skill_ids:
rec = self._resolve_target_skill_record(target_id)
if not rec:
logger.warning(f"Target skill not found: {target_id}")
return None
content = self._load_skill_content(rec)
if not content:
logger.warning(f"Cannot load content for skill: {target_id}")
return None
skill_dir = Path(rec.path).parent if rec.path else None
records.append(rec)
resolved_target_ids.append(rec.skill_id)
contents.append(content)
if skill_dir:
dirs.append(skill_dir)
if resolved_target_ids != suggestion.target_skill_ids:
logger.info(
"Resolved evolution targets %s -> %s",
suggestion.target_skill_ids,
resolved_target_ids,
)
suggestion = copy.copy(suggestion)
suggestion.target_skill_ids = resolved_target_ids
# FIX must target exactly one skill
if suggestion.evolution_type == EvolutionType.FIX and len(records) != 1:
logger.warning(
f"FIX requires exactly 1 target, got {len(records)}: "
f"{suggestion.target_skill_ids}"
)
return None
return EvolutionContext(
trigger=EvolutionTrigger.ANALYSIS,
suggestion=suggestion,
skill_records=records,
skill_contents=contents,
skill_dirs=dirs,
source_task_id=analysis.task_id,
recent_analyses=[analysis],
available_tools=self._available_tools,
capture_dir=capture_dir,
)
def _resolve_target_skill_record(self, target: str) -> Optional[SkillRecord]:
"""Resolve an analyzer target to an active SkillRecord.
Analyzer prompts ask for true skill IDs, but models sometimes emit the
human-readable skill name or a path-like directory name. Accept those
forms here so a valid FIX/DERIVED suggestion does not fail before the
evolver can inspect evidence.
"""
value = str(target or "").strip()
if not value:
return None
rec = self._store.load_record(value)
if rec:
return rec
active_records = self._store.load_active()
by_name = [
r for r in active_records.values()
if r.name == value or Path(r.path).parent.name == value
]
if len(by_name) == 1:
return by_name[0]
if len(by_name) > 1:
logger.warning(
"Ambiguous evolution target %r matched active skills: %s",
value,
[r.skill_id for r in by_name],
)
return None
return None
def _load_skill_content(self, record: SkillRecord) -> str:
"""Load SKILL.md content from disk via registry or direct read."""
# Try registry first (uses cache, keyed by skill_id)
content = self._registry.load_skill_content(record.skill_id)
if content:
return content
# Fallback: read directly from path
if record.path:
p = Path(record.path)
if p.exists():
try:
return p.read_text(encoding="utf-8")
except Exception:
pass
return ""
@staticmethod
def _format_skill_dir_content(skill_dir: Path) -> str:
"""Format all text files in a skill directory for prompt inclusion.
Returns a multi-file listing if there are auxiliary files beyond
SKILL.md, or just the SKILL.md content for single-file skills.
"""
files = collect_skill_snapshot(skill_dir)
if not files:
return ""
# Single-file skill: return just the content
if len(files) == 1 and SKILL_FILENAME in files:
return files[SKILL_FILENAME]
# Multi-file: format as directory listing
parts: list[str] = []
# SKILL.md first
if SKILL_FILENAME in files:
parts.append(f"### File: {SKILL_FILENAME}\n```markdown\n{files[SKILL_FILENAME]}\n```")
for name, content in sorted(files.items()):
if name == SKILL_FILENAME:
continue
parts.append(f"### File: {name}\n```\n{content}\n```")
return "\n\n".join(parts)
@staticmethod
def _format_analysis_context(analyses: List[ExecutionAnalysis]) -> str:
"""Format recent analyses into a concise context block for prompts."""
if not analyses:
return "(no execution history available)"
parts: List[str] = []
for a in analyses[:_ANALYSIS_CONTEXT_MAX]:
completed = "completed" if a.task_completed else "failed"
# Per-skill notes
skill_notes = []
for j in a.skill_judgments:
applied = "applied" if j.skill_applied else "NOT applied"
note = f" - {j.skill_id}: {applied}"
if j.note:
note += f"{j.note[:_ANALYSIS_NOTE_MAX_CHARS]}"
skill_notes.append(note)
# Tool issues
tool_lines = []
for issue in a.tool_issues[:3]:
tool_lines.append(f" - {issue[:200]}")
block = f"### Task: {a.task_id} ({completed})\n"
if a.execution_note:
block += f"{a.execution_note[:_ANALYSIS_NOTE_MAX_CHARS]}\n"
if skill_notes:
block += "Skills:\n" + "\n".join(skill_notes) + "\n"
if tool_lines:
block += "Tool issues:\n" + "\n".join(tool_lines) + "\n"
parts.append(block)
return "\n".join(parts)