ai-sdlc-harness

[HARNESS INTERNAL — do not invoke directly] Implementation specialist, activated exclusively by the ai-sdlc-harness dev-workflow orchestrator during Phase 3 (Development Loop). Never invoke this agent outside the harness workflow.

---

---

[HARNESS INTERNAL — do not invoke directly] Phase 7 PR comment analyser, activated exclusively by the ai-sdlc-harness dev-workflow orchestrator. Read by the reviewer agent when invoked with mode: pr-comment-analysis. Never invoke outside the harness.

[HARNESS INTERNAL — do not invoke directly] Phase 6 pre-PR holistic reviewer, activated exclusively by the ai-sdlc-harness dev-workflow orchestrator. Read by the reviewer agent when invoked with mode: pre-pr. Never invoke outside the harness.

One-time workspace setup. Generates local context files (.claude/context/, git-ignored) that all workflows depend on — repo metadata, coding conventions, local repo paths, language configuration (discovered, not hard-coded), and provider configuration. Supports ADO, Jira, GitLab, GitHub, Zoho, and local-markdown as work item providers, plus GitHub CLI (gh-cli) and GitLab CLI (glab-cli) as no-MCP git providers. Run once before using /dev-workflow or /story-workflow.

Pure aggregator for workflow metrics. Reads timestamp stamps from a tracker and produces `metrics-report.md` (per-workflow) plus appends one row to workspace-level `ai/_metrics-log.csv`. No agent reasoning — deterministic data aggregation only.

Decompose an approved requirements summary into an implementation plan with atomic tasks, Mermaid diagrams, and a task tracker. Use during Phase 2 of the development workflow after requirements are fully understood.

Create a Pull Request / Merge Request after all tasks and tests are approved. Assembles PR summary from the task tracker, validates branch naming, and creates the PR/MR via the configured git provider. Links back to the work item via the work item provider. Supports ADO, GitLab, GitHub, and Jira. Used during Phase 6 of the workflow.

Composes the per-repo static security report (SAST + dependency-vuln + secret-scan). Discovery-driven tool selection from `.claude/context/language-config.md`. Severity normalisation via the per-tool map. Returns 0 (no findings ≥ medium) / 1 (findings present) / 2 (precondition unmet — tool missing).

16 hooks across 8 events