agent-engram

Scaffolds compilable but failing test harnesses for feature and chore tasks

Structured code review using tiered persona subagents, confidence-gated findings, and a merge/dedup pipeline. Use when reviewing code changes before creating a PR, as a build gate, or for standalone review.

On-demand harness health diagnostic. Checks manifest presence, version drift, file integrity, cross-reference validity, MCP tool availability, and template variable residue. Produces a per-category graded health report.

Decomposes a reviewed implementation plan into backlog feature/task/subtask hierarchy

Multi-persona review gate for implementation plans. Validates architectural soundness, scope boundaries, and coding standards compliance before the harvest skill decomposes a plan into work items.

Multi-phase security audit skill. Scans agentic config surfaces and application source for OWASP Top 10 vulnerabilities, STRIDE threats, and config hygiene issues. Produces a scored, graded report persisted to the configured security docs directory (default: docs/security).

Manages the full PR lifecycle: creation, review comment handling, CI remediation, and user-approved merge

Scan .copilot-tracking/ for stale or oversized tracking artifacts, produce compacted summaries, and archive originals. Use when tracking files accumulate beyond thresholds that degrade agent context quality.

Document a recently solved problem to compound institutional knowledge. Captures solutions in .backlog/compound/ with searchable YAML frontmatter for the learnings-researcher. Use after fixing bugs, resolving build issues, or discovering gotchas.

Transform feature descriptions or requirements into structured implementation plans grounded in repo patterns and research. Use when the user says 'plan this', 'create a plan', 'how should we build', 'break this down', or when a brainstorm requirements document is ready for technical planning.

Usage: Build feature {task-id} with harness {harness-cmd}. Implements a requested feature by continuously looping a fast worker agent against a strict, compiling, but failing test harness until success is achieved.

Usage: Fix CI. Detects CI pipeline failures and GitHub Copilot review comments on the current branch's PR, reproduces and fixes errors locally, addresses review comments, runs all CI gates, then pushes and polls until the pipeline passes and comments are resolved.