Part of the Overpowered skill suite. Find and compare relevant historical cases—contracts, proposals, projects, incidents, decisions, designs, or exceptions—to inform a current problem without treating similarity as authority. Use when the question is effectively 'have we dealt with something like this before?' or reusable prior work may exist.

Install

openclaw skills install @raguets/find-precedent

Overpowered / Find Precedent

Core rule

Find comparable prior cases, explain why they are comparable, extract reusable lessons, and make material differences impossible to miss. Never convert precedent into authority.

Procedure

  1. Define the current case and the decision the precedent should inform.
  2. Extract 3–7 comparison dimensions that actually matter: customer type, requirement, risk, contract shape, architecture, scale, jurisdiction, outcome, etc.
  3. Search the most appropriate historical source. If source selection is unclear, use know-enough first.
  4. Rank candidate precedents by decision-relevant similarity, not wording similarity alone.
  5. For each useful precedent, capture:
    • why it is comparable;
    • decision/action taken;
    • known outcome;
    • material differences;
    • what is reusable;
    • what must not be copied blindly.
  6. Synthesize patterns across cases. Preserve outliers when they matter.

Output contract

text
Current decision: ...

Precedent 1
- Similar because: ...
- What was done: ...
- Outcome: ...
- Material differences: ...
- Reusable: ...
- Do not copy blindly: ...

Pattern across precedents: ...
Limits of analogy: ...

Gotchas

  • Semantic similarity is only a candidate-generation signal.
  • A signed historical contract does not define current legal policy.
  • Old proposals may contain stale features, prices, dates, staffing, or commitments.
  • Do not hide cases that contradict the preferred answer.
  • If no sufficiently comparable precedent exists, say so.

Stop condition

Stop when enough comparable cases exist to identify a stable pattern or when further search is unlikely to improve the analogy.