Map Seed
Plans around fixed constraints, runs a fast throwaway prototype, retros what worked versus what didn't, then restarts carrying the learnings instead of polishing a doomed first run. Use when the user wants to prototype or de-risk a design before committing, work out the kinks under fixed or awkward constraints, or run a quick throwaway then iterate.
No Frontmatter
A Claude skill from dylanmccavitt/loom.
Double Dash
Use when testing bad hyphen names.
Outer
Use when testing nested skill file detection.
Assembler
Sets a repository up for the Factorio workflow kit or refreshes its envelope by reading the repo and generating the repo-local envelope every other kit skill binds to, plus the harness wiring and repo-specific skills and agents. Use when setting up a repo for the kit or refreshing its tracker team/project/label map, domain glossary, build/test/lint commands, PR/issue/doc templates, or repo-specif…
Space Age
Coordinates delivery beyond a single repo or environment — promoting an artifact through a CI/CD pipeline planet by planet and landing cross-repo changes in dependency order, reusing the per-hop launch gates. Use when work crosses one repo or one environment: CI/CD pipelines, releasing/promoting through environments, or coordinating a change across multiple repos/services.
Leading
Use when testing bad hyphen names.
Duplicate Skill
Use when testing duplicate skill names in fixture B.
Trailing
Use when testing bad hyphen names.
Secret Skill
Use when testing token detection in version-controlled skills.
Biters
Adversarial reviewer that attacks a change before merge — probing for correctness bugs, regressions, maintainability rot, scope creep, and missing tests by default, with security, minimal-diff, and workflow-drift passes available as lenses. Use when the user wants a change reviewed adversarially, risks found before merge, or a lens-focused review pass (correctness, security, minimal-diff, drift).
Roboports
The implement coordinator. Runs one tracked issue end-to-end as code — one issue to one branch/worktree to one PR — with localized subagent fanout and a minimal diff, and covers behavior-preserving refactors and measured performance work through lenses. Use when the user asks to start, continue, or ship one tracked issue, refactor without changing behavior, or optimize a proven bottleneck.
Nested
Use when testing nested skill file detection.
<skill Name>
<One-line role summary with concrete triggers, synonyms, and near-miss routing hints. Use when <specific user intent or packet condition>.>
Blueprint
The shape owner. Turns current context into a PRD/spec with acceptance criteria, non-goals, and a proof plan, and covers issue decomposition, architecture seams, research spikes, and tracker triage through lenses. Owns the kit's reusable PR, issue, project-doc, and PRD templates. Use for any shape-mode work - specs, breaking work into issues, planning structure, resolving unknowns, or triaging in…
Bad Version
Bad version fixture. Use when validator tests reject invalid metadata versions.
Duplicate Skill
Use when testing duplicate skill names in fixture A.
Repair Pack
Use when fixing exactly one concrete review or proof finding from a fresh compact packet.
Rocket Launch
Ship a ready change off-planet by enforcing the launch gates, merging the PR, and letting the bridge close its tracked issue. Use when a change is ready to ship — merge the PR, run the review gate, and close out the tracked issue; not for opening a draft or work that is not ready (that stays with `roboports`).
Fine Grained Pat
Use when testing fine-grained GitHub token detection.
Belt
Handoff carrier that moves durable context between agents or threads with concise state, proof, risks, and next actions, covering handoff, thread-control, and resume flows through lenses. Use when writing a handoff, deciding whether to continue or start a new thread, or resuming a repo task from durable state.
Handoff
Use when testing collision with an existing global handoff skill.
Bad Leading Zero Dotted Prerelease
Bad dotted prerelease fixture. Use when validator tests reject leading-zero numeric prerelease identifiers after dots.
Prospect
Lands a brand-new idea, feature, or initiative as tracked planning work and creates its tracker home before any spec or issues exist. Use when the user is starting a new idea, feature, or initiative from scratch and wants it captured as planning work before a spec or issues exist.
Alpha One Prerelease
Valid prerelease fixture. Use when validator tests accept dotted alphanumeric prerelease identifiers.
Zero Alpha Prerelease
Valid alphanumeric prerelease fixture. Use when validator tests accept leading digits in alphanumeric prerelease identifiers.
Bad Long Description
Use when xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx…
Lab
Proof specialist that runs proof-only validation and records behavior evidence without expanding scope, covering command, UI, and smoke proof through lenses. Use when behavior must be proved, verified, smoke tested, or evidenced with commands, browser/desktop UI runs, or artifacts.
Missing Version
Missing version fixture. Use when validator tests reject absent metadata versions.
Zero Prerelease
Valid zero prerelease fixture. Use when validator tests accept zero numeric prerelease identifiers.
Bad Leading Zero Prerelease
Bad prerelease fixture. Use when validator tests reject leading-zero numeric prerelease identifiers.
Good Skill
Use when validating a known-good version-controlled skill fixture.
Bad Name
Use when validating rejection of skill names outside agentskills.io charset rules.