claude

agentstrap

Workspace bootstrapper: idempotent scaffolding + sanity check, automatic cross-device handoffs, audit swarm, release pipeline.

1 on GitHubby mateim4v0.4.0MITupdated 15 days ago
/

Plugin agentstrap from marketplace agentstrap-v2 in mateim4/agentstrap-v2. Commands read from the repo's manifest — 15 days ago.

This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.

What it installs

Skills (9)

  • audit

    Run the AgentStrap audit swarm over the codebase — fan out specialist reviewers in parallel, then consolidate into a deduped P0–P3 report. Use when asked to audit, review, or do a quality/security pass on a project with code.

  • bootstrap

    Bootstrap or sanity-check an AgentStrap workspace in this project — idempotent and non-destructive. Detects greenfield vs existing-docs vs already-stamped, reports gaps, and adds only what is missing while conforming to existing conventions. Run when setting up AgentStrap on a project or re-checking

  • consolidate-findings

    Reduce step for the audit swarm — dedupe findings from multiple reviewers, assign stable IDs, normalize to P0–P3, and group into prioritized batches with a dependency note. Used by /agentstrap:audit and /agentstrap:security-audit.

  • decision

    Record an architectural/product decision in the project's Decisions Log (append-only, newest first). Use when a choice has been made and should be captured durably.

  • handoff

    Force a full handoff refresh now — write a thorough narrative of where work stands into HANDOFF.md so the next session (or device) resumes cleanly. Use when wrapping up, before switching machines, or when asked to checkpoint.

  • open-question

    Add or resolve an entry in the project's Open Questions register. Resolving migrates it to the Decisions Log. Use to track unresolved forks so nothing falls through.

  • release

    Run the AgentStrap release pipeline — bump all version files, finalize the changelog, build, gate on tests, tag, and write checksums. Never auto-pushes. Use when cutting a release.

  • security-audit

    Focused security audit — runs the security + adversarial reviewers with STRIDE and deployment-context severity, then consolidates. Use for a security-only pass.

  • working-rules

    The AgentStrap PM-first collaboration contract — how Claude should drive an AgentStrap project. Load when starting work, planning, making architectural choices, or when unsure how to collaborate on an AgentStrap-bootstrapped project.

Subagents (9)

  • audit-adversarial

    Read-only red-team auditor. Chains individual weaknesses into concrete end-to-end abuse scenarios. Dispatched by /agentstrap:audit.

  • audit-architecture

    Read-only architecture auditor. Reviews module boundaries, coupling, data flow, and scalability of the design. Dispatched by /agentstrap:audit.

  • audit-code-quality

    Read-only code-quality auditor. Reviews correctness, readability, error handling, dead code, and duplication. Dispatched by /agentstrap:audit.

  • audit-documentation

    Read-only documentation auditor. Checks README/setup accuracy, API docs, changelog completeness, and misleading comments. Dispatched by /agentstrap:audit.

  • audit-performance

    Read-only performance auditor. Finds hot-path inefficiencies, N+1 queries, render waste, bundle bloat, and leaks. Dispatched by /agentstrap:audit.

  • audit-security

    Read-only security auditor. Finds vulnerabilities and tags them with STRIDE + deployment-context severity. Dispatched by /agentstrap:audit and /agentstrap:security-audit.

  • audit-testing

    Read-only test-quality auditor. Finds coverage gaps, missing edge cases, flaky patterns, and weak assertions. Dispatched by /agentstrap:audit.

  • audit-ux-accessibility

    Read-only UX/accessibility auditor. Checks WCAG 2.2 AA, keyboard operability, loading/empty/error states, and Core Web Vitals. Dispatched by /agentstrap:audit.

  • research-delegate

    Worker for multi-query research and codebase exploration so the main thread stays context-clean (working rule 9). Use for web/competitor/regulatory scans or any exploration beyond ~3 queries. Returns only distilled findings.

Hooks (3)

  • SessionStart

    bash "${CLAUDE_PLUGIN_ROOT}/scripts/session-start.sh"

  • Stop

    bash "${CLAUDE_PLUGIN_ROOT}/scripts/on-stop.sh"

  • SessionEnd

    bash "${CLAUDE_PLUGIN_ROOT}/scripts/session-end.sh"

View source on GitHub

Maintain agentstrap?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex pluginslisted