Agent Claude Code
Runs one phase or slice end to end in its own context and returns exactly one summary line, so the orchestrator above stays lean.
not rated 5 1mo ago A 31 tokens
original MIT archived
editor
02Agent Claude Code
Sole source mutator in fix loops. Fixes one module's punch-list items at one fix level, lands its changes, and returns its changelog lines instead of editing the changelog.
not rated 5 1mo ago A 41 tokens
original MIT archived
planner
03Agent Claude Code
Decomposes one chunk of the plan into a forward-only DAG in its own context, writes the full plan to disk, and returns a compact summary.
not rated 5 1mo ago A 33 tokens
original MIT archived
Agent Claude Code
Runs one complete review round (lanes, reviewer fan-out, triage, editor waves, integration, verification) in its own bounded context and returns exactly one summary line.
not rated 5 1mo ago A 40 tokens
original MIT archived
reviewer
05Agent Claude Code
Read-only reviewer for one review dimension on one module shard. Returns its findings as an EDN vector, or exactly NO FINDINGS.
not rated 5 1mo ago A 30 tokens
original MIT archived
ui-designer
06Agent Claude Code
Designs a UI surface before it is built and reviews UI surfaces against the design system. Produces design specs and view-spec data, not runtime code. Active only when the descriptor opts in.
not rated 5 1mo ago A 43 tokens
original MIT archived
verifier
07Agent Claude Code
Runs the deterministic verification lanes on a change and reports pass, skip, or fail with the first error only. Cheap, bash-heavy, no judgment calls.
not rated 5 1mo ago A 34 tokens
original MIT archived
writer
08Agent Claude Code
Writes new code for one unit of work, in its own jj workspace or inline in the working copy, following the write- or write-tests recipe named in its dispatch.
not rated 5 1mo ago A 39 tokens
original MIT archived
Skill Claude Code
Generate a check- recipe for a bug class the active set misses.
not rated 5 1mo ago A 20 tokens
original MIT archived
add-language
10Skill Claude Code
Generate a write- recipe for a language outside the curated four.
not rated 5 1mo ago A 18 tokens
original MIT archived
add-tech
11Skill Claude Code
Add a non-language concern, a lane entry, a spine task, or a hook.
not rated 5 1mo ago A 20 tokens
original MIT archived
advance-plan
12Skill Claude Code
Drive a chunk of the plan to completion unattended, phase by phase.
not rated 5 1mo ago A 16 tokens
original MIT archived
Skill Claude Code
Trace the causal chain from trigger to failure.
not rated 5 1mo ago A 14 tokens
original MIT archived
Skill Claude Code
Fix the punch-list findings assigned to you, one module at one level, smallest sufficient edit, verify then land. Run by editor agents.
not rated 5 1mo ago A 32 tokens
original MIT archived
assess-data
15Skill Claude Code
Plan schema changes, migrations, and backfills.
not rated 5 1mo ago A 13 tokens
original MIT archived
Skill Claude Code
Plan logging, metrics, and tracing for a feature.
not rated 5 1mo ago A 16 tokens
original MIT archived
assess-risk
17Skill Claude Code
Assess pre-deployment risk with rollback plans.
not rated 5 1mo ago A 12 tokens
original MIT archived
Skill Claude Code
Define deployment strategy and verification.
not rated 5 1mo ago A 10 tokens
original MIT archived
Skill Claude Code
Define the testing strategy for a feature.
not rated 5 1mo ago A 11 tokens
original MIT archived
audit-code
20Skill Claude Code
Review and fix any scope, round after round until a dry round.
not rated 5 1mo ago A 16 tokens
original MIT archived
Skill Claude Code
Detect the stack, write the descriptor, materialize recipes and scaffolding.
not rated 5 1mo ago A 17 tokens
original MIT archived
Skill Claude Code
Capture a correction or interaction preference mid-session at near-zero cost, route it to the right home, and return to the work.
not rated 5 1mo ago A 30 tokens
original MIT archived
Skill Claude Code
Review dimension for reader experience, jargon, and pacing over prose and docs. Invoked by reviewer agents over a prose shard when the descriptor activates :clarity.
not rated 5 1mo ago A 37 tokens
original MIT archived
Skill Claude Code
Review dimension for divergence from the spec, the ADRs, and the language standard. Invoked by reviewer agents over one module shard.
not rated 5 1mo ago A 31 tokens
original MIT archived
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: