core/capabilities/review/spec-review/SKILL.md
Adversarial review verifying implementation matches its specification — checks completeness (nothing missing) and precision (nothing extra). Distrusts the implementer's self-report. Use after implementation, or when the user says "check against spec", "does this match requirements", or "verify the implementation".
npx skillsauth add xoai/sage spec-reviewInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Verify that the implementation matches its specification — nothing more, nothing less.
Core Principle: Do NOT trust the implementer's report. The implementer may have finished too quickly, missed requirements, added unrequested features, or described what they intended rather than what they built. Verify EVERYTHING independently.
After every task implementation, as part of the quality gate sequence. This is Gate 1 of 5 — it runs first because all other gates are meaningless if the code doesn't even match what was requested.
Read the COMPLETE task specification. Understand:
Read the ACTUAL code that was written. Not the implementer's summary. Not the commit message. The code itself.
For each requirement in the specification:
For each piece of code in the implementation:
GATE: spec-compliance
RESULT: PASS | FAIL
REQUIREMENTS CHECK:
✓ [requirement 1] — Implemented and tested
✓ [requirement 2] — Implemented and tested
✗ [requirement 3] — MISSING: not implemented
✗ [requirement 4] — PARTIAL: implemented but not tested
SCOPE CHECK:
✓ No unrequested features added
— OR —
✗ EXTRA: [description of unrequested code]
ACTION: none | fix-and-retry | escalate-to-human
Assume the implementer:
Your job is to be the skeptic. If you can't find evidence that a requirement is implemented and tested, it isn't — regardless of what anyone claims.
development
Branch-per-initiative git discipline for all delivery workflows. Defines branch naming by workflow, the propose-confirm creation protocol, dirty-tree and detached-HEAD handling, the always user-gated merge protocol, worktree support for parallel sessions, and abandonment cleanup. Activates only in git repositories — silently inactive everywhere else. Use when starting /build, /fix, /architect, or /build-x at Standard+ scope, when resuming an initiative, when offering a merge at a completion checkpoint, or when the user wants a second concurrent initiative.
development
Drives task-by-task execution from an approved plan with quality gates between each task. Reads the plan, finds the next incomplete task, dispatches implementation, validates, updates progress, and continues. Use after a plan is approved and the user says "go", "start building", "execute the plan", or "implement the feature".
testing
Preserves and restores context across agent sessions using plan file checkboxes as source of truth. Use when starting a new session, resuming previous work, ending a session, or when the user says "continue from last time", "what was I doing", or "save progress".
tools
Captures agent mistakes, corrections, and discovered gotchas so they are not repeated. Use when: (1) a command or operation fails unexpectedly, (2) the user corrects the agent, (3) the agent discovers non-obvious behavior through debugging, (4) an API or tool behaves differently than expected, (5) a better approach is found for a recurring task. Also searches past learnings before starting tasks to avoid known pitfalls. Activate alongside the sage-memory skill — they share the same MCP backend but serve different purposes (sage-memory = codebase knowledge, sage-self-learning = agent mistakes and gotchas).