skill-evidence

Before proposing work here

This crate is depended on by playbench, mundifold, and what-we-bring-home, which between them hold over a thousand append-only recorded events that no release can regenerate. Read docs/principles/ before proposing anything that changes the public API, the installed skill packages, the published schemas, or the shape of a recorded event.

Two things that catch people out:

• Cargo SemVer protects the Rust API and nothing else. Adding an optional field to a recorded event is a compatible change by every Cargo rule and can still invalidate a consumer's history. See docs/principles/consumer-contract.md.
• Install never removes. Retiring or renaming an installed package leaves it in every consumer until the consumer deliberately runs skills evidence withdraw, or follows the release note's exact manual fallback when pinned below the withdrawal version. See docs/releasing.md.

This repository has no value stream of its own — everything here is enabling work for a consumer. Its own issue count is not progress.

Agent skills

Skill prose writes skill names as /name — Claude Code's trigger syntax. Codex invokes the same skill as $name. Read /name as the skill called name, and when you tell the user to run one, use the prefix of the harness you are in.

Issue tracker

Issues live as GitHub issues in joeloverbeck/skill-evidence, driven by the gh CLI; external PRs are also a triage surface. See docs/agents/issue-tracker.md.

Triage labels

The five canonical strings, unmodified: needs-triage, needs-info, ready-for-agent, ready-for-human, wontfix. See docs/agents/triage-labels.md.

Domain docs

Single-context: CONTEXT.md and docs/adr/ at the repo root. See docs/agents/domain.md.