AI-agent integration
How well the repository supports AI coding agents — of any vendor — as first-class contributors. No external standard exists for this, so it is a deliberate equal-weight presence checklist (documented in MATURITY.md): each item is one point because ranking them would be judgement, not measurement.
5.0 / 510 of 10 weighted points earned — snapshot of 2026-08-04
agents-mdweight 1 ptearned 1 / 1
Measures: AGENTS.md (cross-vendor source of truth)
Why this weight: The cross-vendor instruction file every agent can read — the foundation of the checklist; one point like every item in this deliberately unranked dimension.
claude-imports-agentsweight 1 ptearned 1 / 1
Measures: CLAUDE.md points at AGENTS.md, not a copy
Why this weight: Verifies the per-tool file is a pointer, not a fork — duplicated instructions drift. Equal weight by the checklist principle.
spec-prompt-hookweight 1 ptearned 1 / 1
Measures: a prompt hook injects the spec workflow
Why this weight: Puts the spec-driven workflow in front of the agent on every prompt, instead of relying on it re-reading docs. Equal weight by the checklist principle.
precommit-agent-hookweight 1 ptearned 1 / 1
Measures: an agent pre-commit hook reaches the shared gate
Why this weight: Fast in-session feedback that delegates to the vendor-neutral git hook, so the check logic lives in one place. Equal weight by the checklist principle.
coverage-agent-hookweight 1 ptearned 1 / 1
Measures: an agent hook checks coverage after edits
Why this weight: Surfaces coverage regressions right after an edit instead of at commit time. Equal weight by the checklist principle.
session-bootstrapweight 1 ptearned 1 / 1
Measures: cold-session bootstrap (script + hook)
Why this weight: A cold container that cannot install dependencies wastes every subsequent agent action; bootstrap makes sessions productive from the first prompt. Equal weight by the checklist principle.
permissions-allowlistweight 1 ptearned 1 / 1
Measures: a permissions allowlist reduces prompts
Why this weight: Pre-approving known-safe commands keeps agents autonomous without widening what they may do. Equal weight by the checklist principle.
portable-mcpweight 1 ptearned 1 / 1
Measures: project-scoped MCP config (.mcp.json)
Why this weight: Tool access declared once via the open protocol works for any MCP-capable agent, instead of per-vendor integrations. Equal weight by the checklist principle.
pr-template-idsweight 1 ptearned 1 / 1
Measures: the PR template asks for requirement IDs
Why this weight: Makes requirement traceability part of the review surface for human and agent PRs alike. Equal weight by the checklist principle.
machine-readable-stateweight 1 ptearned 1 / 1
Measures: build state is machine-readable (traceability.json)
Why this weight: Agents plan far better against structured state than prose; the matrix is the machine-readable ground truth. Equal weight by the checklist principle.