[codexstar69]

Purpose

Bring every current agent-facing surface in line with the measurable precision-first Bug Hunter protocol, while removing historical prompt residue, contradictory runtime guidance, and publication paths that did not enforce the full quality contract.

What changed

• rewrote the top-level SKILL.md as a compact, progressively loaded control plane rather than a repeated knowledge dump
• aligned scan-only and single-pass defaults; --loop is explicit and mutation/commit authority remains separately gated
• updated README, onboarding, installation, usage, CLI, architecture, troubleshooting, contributor, security, bundled-skill, LLM, dispatch, and subagent-wrapper guidance
• made current-source versus published-package provenance explicit so documentation does not imply npm latest necessarily contains the exact GitHub main implementation
• replaced stale eval assumptions with 39 current scenarios covering adaptive policy, bounded retrieval, content-addressed evidence reuse, hybrid verification, source integrity, truthful coverage, security routing, and immutable Fixer scope
• added permanent freshness regressions for legacy task prompts, old artifact names, historical loop-default claims, missing measurable artifacts, oversized skill instructions, source-release ambiguity, and incomplete publication gates
• updated package metadata and changelog without presenting an untagged source version as an already-published release
• removed temporary self-modifying workflow carriers before publication
• hardened the release-only npm workflow so publication now requires generated assets, the full test suite, benchmark quality gate, runtime preflight, package inventory, exact tag/main provenance, and OIDC trusted publishing

Final exact-head verification

Verified on head 015913ac952f862b49bc423ac36d4ea44752b93d:

• generated schema validators and compatibility prompts: passed
• complete Node.js test suite: 217 passed, 0 failed on Node.js 22 and Node.js 24
• benchmark quality gate: passed
• runtime preflight: passed
• package inventory: passed (97 files, no forbidden entries or errors)
• executable-bit, --version, and --help checks: passed
• sealed precision protocol gate and sealed-check hash: passed

Bundled deterministic benchmark fixture on this head:

• precision: 1.00
• recall: 1.00
• F1: 1.00
• severity-weighted recall: 1.00
• repeat stability: 1.00
• false positives per KLOC: 0
• median tokens per true positive: 12,090
• p95 duration: 61.3s
• expected calibration error: approximately 0.0483

These fixture figures validate the shipped benchmark and quality-gate machinery; they are not presented as independent proof of universal superiority across every repository, language, model, or runtime.

Safety and publication state

• no release, npm publication, deployment, or merge is performed by this PR
• scan-only and single-pass remain the defaults
• required verification failure still blocks Fixer authorization
• source integrity, assigned-file scope, truthful per-file coverage, resume identity, and immutable Fixer scope remain fail-closed
• the PR contains only the 23 intended runtime/documentation/test files; temporary implementation workflows are absent

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.