agentsclimarketplace
Sign in

Spec synthesis review

Skill mwarger/forge/skills/spec-synthesis-review

Reduce typed sub-agent outputs into canon and verify the subject spec before readiness promotion. Use this when you need deterministic merges, provenance review, contradiction handling, verification loops, section drafting, or final review passes across the spec.From its SKILL.md

Install
npx -y skills add mwarger/forge --skill spec-synthesis-review

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.5 KB, as published. Nobody here has run it

Use this after a scoring pass or when draft sections need promotion into canon.

Reducer rules

  • The lead agent is the only reducer.
  • Apply sub-agent results only when parent_revision matches canon.
  • If the result is stale, rebase, revalidate, or discard.

Merge order:

  1. evidence / claims
  2. contradictions
  3. ontology rescore
  4. section patch
  5. readiness gate

Required review passes

  1. completeness review
  2. contradiction review
  3. provenance review
  4. implementability review
  5. terminology review

Verification loop

Before readiness promotion, run:

  • completeness check against requested deliverables
  • grounding check from canonical text to claim ids
  • empty-result recovery for suspiciously thin evidence
  • section/schema format check
  • missing-context gate
  • action-safety gate for mutation or publication steps
  • stable-id check for questions, assumptions, decisions, and major claims
  • evidence-ref check for every major canon claim
  • terminology check against UBIQUITOUS-LANGUAGE.md at the project root
  • blocker-state check against the orchestrator readiness contract

If any check fails, return to:

  • evidence gathering
  • targeted questions
  • contradiction review
  • or redrafting

Root specs index update

After the subject spec reaches a stable draft, create or update the row for the current subject inside the managed block in specs/README.md.

Managed block markers:

  • <!-- forge:spec-index:start -->
  • <!-- forge:spec-index:end -->

Preserve prose outside the managed block.

The row should include:

  • Spec
  • Target
  • Purpose
  • Planning Status
  • Handoff
  • Keywords
  • Artifacts

Rules:

  • Target should point at the implementation directory or code path when it can be inferred, otherwise use —
  • Purpose should be one short implementation-oriented sentence
  • update an existing row for the subject instead of appending duplicates
  • the root index is navigational; the subject spec remains the canonical implementation artifact

Output

Update:

  • specs/<subject>.md
  • specs/README.md
  • review-report.md
  • claim-ledger.jsonl
  • contradiction-log.md
  • decision-log.md
  • branch-registry.json

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.