agent-skills-toolchain

Launch the agent-monitor — a live dashboard of Claude Code subagents, including agents running inside Workflows.

Answer a quick side question without interrupting or losing context from the current task. Resume work automatically after answering.

Invoke the metaswarm brainstorming extension skill, which wraps `superpowers:brainstorming` with metaswarm's design review gate handoff.

Incrementally fix build and type errors with minimal, safe changes.

Create or verify a checkpoint in your workflow.

ADVISORY, evidence-citing Agent-as-a-Judge. Decomposes a task into a YES/NO Definition-of-Done checklist, cites a ledger-event or evidence-hash (chain-of-custody) per item, execute-to-confirms every finding it would call blocking, and scores each item against the ICD-203 nine analytic standards. Runs AFTER the deterministic verification cluster; its opinion NEVER gates or grants "done" (COUNCIL C3) — only a re-run executable receipt it cites could ever block, and that path is DEFERRED INDEFINITELY this wave, so output is 100% advisory. Imports passk-stats (F5) for kappa / e-value calibration against the owner's claude-reflect corrections; stays advisory while kappa < 0.6 (the honest prior) and auto-demotes on kappa decay. Invoke to annotate residual SUBJECTIVE DoD items the deterministic gates cannot judge. NOT a done-gate, NOT terminal authority, NEVER edits source, NEVER blocks, NEVER nags.

Software architecture specialist for system design, scalability, and technical decision-making. Use PROACTIVELY when planning new features, refactoring large systems, or making architectural decisions.

Build and TypeScript error resolution specialist. Use PROACTIVELY when build fails or type errors occur. Fixes build/type errors only with minimal diffs, no architectural edits. Focuses on getting the build green quickly.

Personal communication chief of staff that triages email, Slack, LINE, and Messenger. Classifies messages into 4 tiers (skip/info_only/meeting_info/action_required), generates draft replies, and enforces post-send follow-through via hooks. Use when managing multi-channel communication workflows.

Expert code review specialist. Proactively reviews code for quality, security, and maintainability. Use immediately after writing or modifying code. MUST BE USED for all code changes.

Use when you need to run "the smallest test that could still catch this change" instead of blindly running the whole suite (or nothing). Provides an invocable L0 to L3 ladder - L0 lint the edited file, L1 impacted tests via deterministic selectors, L2 full suite via longrun, L3 e2e - so a "tests pass" claim is backed by a proportionate check. Never ML "predictive test selection".

Use when repeated code edits are triggering a full test-suite rerun and burning tokens/time, or before any commit. Splits verification into a cheap INNER loop per edit (only affected/related tests plus type-check and lint on touched files) and a single OUTER loop at the commit boundary (full suite plus coverage plus held-out E2E/golden-journey plus a scoped mutation check).

Use when applying binary templates (.bt), running 010 Editor scripts (.1sc), or performing hex editing with SweetScape 010 Editor. Supports command-line batch processing of binary files with templates and scripts.

Use at the "done" boundary of an app change, before a PR or a "fixed / it works now" claim. Dispatches an INDEPENDENT cross-model reviewer (Codex/Gemini CLI, else fresh Claude) that adversarially judges whether the real end-to-end journey works or is green-but-broken, over the diff + golden-journeys + on-disk evidence. Returns a SHIP/BLOCK verdict. NOT for code review (use code-review).

Advisory, evidence-citing Agent-as-a-Judge. Use for a second opinion on whether a completion meets its Definition of Done AFTER the deterministic verification cluster (typed receipts, fail-first, execution-consensus, golden-journey oracle) ran - decomposes the DoD into falsifiable YES/NO items, cites evidence per item, scores on ICD-203. 100% ADVISORY/SHADOW, never blocks. NOT a done-gate.

26 hooks across 7 events