moonweave
UserLocal Claude Code & Codex agent skills for evidence-gated decisions, measurable build-direction forks, mid-session drift audits, and an honest done-gate (useful-done, not just built-done).
Categories
Indexed Skills (6)
compass
User-invoked mid-session audit skill. Detects (1) drift between current task and original session intent via transcript analysis, and (2) codebase rot — git uncommitted accumulation, test/lint state from cached artifacts, file bloat, TODO accumulation, circular dependencies, module boundary degradation. Manually triggered via `/compass` only — never auto-invoked. Use when the user types `/compass` optionally followed by a baseline intent string. Without arguments, derive baseline via cascade — most recent spec (24h), then task-init heuristic, then explicit user prompt. Never use the literal first user message of the transcript as baseline.
decide
User-invoked research skill. Investigates current industry best practice via parallel WebSearch + context7 + GitHub adoption signals, then either decides autonomously (reversible actions) or escalates with factual-context questions (contested or information-requiring decisions). Manually triggered via `/decide` only — never auto-invoked. Use when the user types `/decide` optionally followed by a question. Without arguments, extract the pending decision from the prior assistant message.
done-gate
A completion-boundary gate that distinguishes built-done (tests pass) from useful-done (a real consumer gets real value) at the moment work is declared finished. Manually triggered via `/done-gate` only — never auto-invoked. Use when the user types `/done-gate` after declaring work done.
anneal
A measurement-first decision discipline for choosing HOW to build something. Before pouring effort into one approach, operationalize a measurable fitness, sketch a few candidate directions, cheaply score each by that fitness, and pick the winner by the number — not by taste, and not by polish. Lightweight and inline: it is a discipline, not a multi-agent engine. Manually triggered via `/anneal` only — never auto-invoked. Use when the user types `/anneal`, typically at a fork where the best approach is not obvious.
anneal
A measurement-first decision discipline for choosing HOW to build something. Before pouring effort into one approach, operationalize a measurable fitness, sketch a few candidate directions, cheaply score each by that fitness, and pick the winner by the number — not by taste, and not by polish. Lightweight and inline: it is a discipline, not a multi-agent engine. Manually triggered via `/anneal` only — never auto-invoked. Use when the user types `/anneal`, typically at a fork where the best approach is not obvious.
ref-verify
Prevents citation hallucination in academic writing. Invoke when: finding papers to support a specific claim; verifying/checking/auditing existing citations or DOIs; confirming whether a paper actually says what the user claims it says ('is that what the paper says?', 'did they actually show X?'); adding a citation by describing a paper ('add a citation for the paper where X'); running a pre-submission reference sweep. Do NOT invoke for: formatting references in APA/IEEE style, general topic explanations, citation style questions, or prose editing. Selects Quick Screen (seconds per paper) or Full Audit (abstract fetch + claim check) automatically.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.