engram-skill-pack v0.2.0 — Structural Bootstrap

Three active skills now. The hero (engram-mistakes) is joined by two skills that bootstrap repo understanding before the agent makes its first edit.

Skills shipping in v0.2.0

Skill	What it does
engram-mistakes	Surface bi-temporal mistakes — auto-fires PreToolUse on Edit/Write/Bash for code files. The rave moment.
engram-query	Natural-language structural queries — replaces ad-hoc Grep with AST-aware lookups at a fraction of the token cost.
engram-gods	Bootstrap repo understanding — top-N entities by importance score. Auto-fires on "where do I start" / "what's important" questions.

Install

npx @anthropic-ai/claude-code add nickcirv/engram-skill-pack

Requires engramx >= 4.0.0 (peer dependency, optional). Install once:

npm i -g engramx
cd ~/your-project && engram init   # auto-installs the Sentinel hook

Quality bar

• All three SKILL.md files follow the full Layer 1-7 AAA structure (control dials + core principles + decision tree + variants + quality gate)
• Each skill has 5 reference files (ranking/score, output-format, trigger-patterns, anti-patterns with ≥7 numbered items, evaluation with ≥4 scoring dimensions)
• 36/36 tests passing (manifest schema + skill structure + frontmatter validation + AAA gate)
• 115 KB unpacked size (clean payload, build artifacts excluded)
• Zero runtime dependencies. Apache-2.0.

What's queued

• v0.3.0 — engram-gen + engram-learn (interactive 4-field bi-temporal capture). Requires engramx learn --interactive to land in engramx core first.
• v0.4.0 — AAA-elevation pass on all 5 skills (full Layer 1-7 cross-validated).
• v1.0.0 — Marketplace launch with comparison-benchmark page (engramx-bench.dev) and Show HN moment.

See CHANGELOG.md for full details.

engram-skill-pack v0.1.0 — Memory Preview

Initial release. Hero skill engram-mistakes only — full AAA Layer 1-7 structure (control dials + core principles + decision tree + invocation variants + quality gate), 5 reference files (bi-temporal model + output format + trigger patterns + anti-patterns + evaluation), 24 tests passing.

Superseded by v0.2.0 within hours (same day) — v0.2.0 adds engram-query + engram-gods. Tagged here for chronological completeness.