// Use when implementing or updating Searchlite features that need pre-review hardening via per-feature invariant matrix maintenance, adversarial/regression coverage updates, and full quality-gate verification.
| name | feature-hardening |
| description | Use when implementing or updating Searchlite features that need pre-review hardening via per-feature invariant matrix maintenance, adversarial/regression coverage updates, and full quality-gate verification. |
Initialize a feature matrix once:
python3 .codex/skills/feature-hardening/scripts/init_feature_hardening.py --feature <feature>Refresh matrix metadata from the current diff each implementation cycle:
python3 .codex/skills/feature-hardening/scripts/update_feature_matrix.py --feature <feature> --base origin/mainExpand the matrix with feature-specific invariants and adversarial cases:
Run the hardening quality gate before requesting review:
.codex/skills/feature-hardening/scripts/run_feature_hardening.sh --feature <feature> --base origin/main --bench-if-sensitivePost a short hardening summary in the PR:
Install an optional pre-push hook:
.codex/skills/feature-hardening/scripts/install_pre_push_hook.shFor CI integration, use the snippet in:
.codex/skills/feature-hardening/references/ci-snippet.mdMatrix path:
docs/feature-hardening/<feature>/matrix.mdThe matrix must include:
Send local completion notifications to the Agent Notifications app via POST http://127.0.0.1:60766/agent/notify after finishing a task; include title, content, and agent details so the desktop app shows a system notification.
Send local completion notifications to the agent-notifier service via POST http://127.0.0.1:60766/agent/notify after finishing a task; include title, content, and agent details so the desktop app shows a system notification.
Structured workflow for discovering, reproducing, and fixing bugs in Searchlite by inspecting real code paths (CLI/HTTP/core), validating durability invariants, and producing regression tests; use when investigating incorrect results, crashes, data loss risks, or “it sometimes fails” reports.
Guardrails for keeping the Searchlite Rust workspace compliant with formatting, lint, build, test, and perf expectations; use when preparing or reviewing changes to ensure they meet project quality bars.
Playbook for diagnosing Searchlite issues across CLI/HTTP, storage, and indexing; use when triaging failures or odd search results.
Style guide for writing Searchlite documentation and examples; use when creating or editing docs/ guides.