scriptoriumadmin/claude-skills
1
0
Fork 0
mirror of https://github.com/alirezarezvani/claude-skills.git synced 2026-10-07 02:58:13 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
claude-skills/docs/commands/cs-deep-research.md
Claude abd9c9d8de
docs(site): generate agent-launcher pages (18th domain) + nav 
generate-docs.py learns the agent-launcher domain (5 hardcoded maps extended);
regenerated docs tree: 343 skill pages / 96 agent pages / 122 command pages
(561 total). mkdocs.yml nav gains the Agent Launcher skill section (7 pages),
4 cs-agent-* agent entries, and 8 /cs:* command entries; all nav targets verified
to exist.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012FwXG6TqCXKZQvF4iD69cv
2026-08-24 17:26:12 +00:00

3.3 KiB
Raw Permalink Blame History

title description
/cs-deep-research — Slash Command for AI Coding Agents /cs:deep-research <question> — Disciplined multi-source investigation for a high-stakes question. Reframes into falsifiable hypotheses, plans, fans. Slash command for Claude Code, Codex CLI, Gemini CLI.

/cs-deep-research

:material-console: Slash Command :material-github: Source

Command: /cs:deep-research <question>

The cs-deep-research persona turns "research this" into an auditable, reusable investigation — the workflow to reach for when getting the answer wrong costs more than the tokens spent getting it right.

When to Run

  • A low-quality answer is expensive: strategy, business plan, report, or article groundwork.
  • Comparing N institutions / products / methods / markets with defensible reasoning.
  • Validating a hypothesis or an irreversible decision against external data.
  • Meta-research: "understand how X works," "map the landscape of Y."

When NOT to Run

  • Quick fact-checks → answer directly.
  • Structured 12-dimension competitor scoring → competitive-teardown.
  • Fast topic overviews where decision risk is low → the research router (/cs:research).

What You Get

  1. Reframe — the question rewritten to the real decision + 2-4 falsifiable hypotheses.
  2. plan.md — genre, sourcing strategy, opposition queries, risk register, stop-criteria.
  3. Parallel search — sub-agents fanned out across channels; each source saved to sources/NN_slug.md with verbatim quotes + Credibility/Recency/Bias scores.
  4. Triangulated synthesis — every thesis backed by >=3 independent, differently-typed sources (or flagged "insufficient evidence"), plus a mandatory adversarial pass.
  5. A reusable folder — sources.csv, findings/, final report, and refresh_targets.md for delta-updates later.

Trigger Phrases (auto-invoke without /cs:)

  • "deep research on [topic]" / "do a deep dive on [topic]"
  • "research this thoroughly / rigorously" / "high-stakes research"
  • "compare [N options] and give me defensible reasoning"
  • "validate this hypothesis with external data"

Discipline

  • No fabricated citations — empty fetch = empty claim.
  • Triangulation mandatory — < 3 independent, differently-typed sources → "insufficient evidence," not fact.
  • Adversarial pass required on medium/deep investigations.
  • Parallel sub-agents — never serial in the search phase.
  • Persist to files — the reuse value is the folder, not the transcript.

Stop Conditions

  • Report written + every thesis triangulated or flagged + refresh_targets.md emitted → done.
  • On an update <slug> run: produce a delta in diffs/ instead of replaying the whole investigation.

Related

Version: 1.0.0