merlin-skills
AGENTS.md1 starsChanged 6 months agoMITView on GitHub
This repository is a skill-first, /goal-first operating system for autonomous web coding. Treat it as a curated workflow layer, not as a grab bag of prompts. For any non-trivial task, load skills/merlin-skills-routing/SKILL.md before selecting implementation skills or creating a Codex /goal. The router must start with a short Spec Kit consultation: inspect .specify/, active specs/<feature>/ artifacts, GOAL.md, task progress, and existing proof before choosing the next skill. The default sequence is: In this sequence, tdd is the implementation discipline for…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,700 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository23 other files it gives its agents.
- brainstormingSkill
- create-goalSkill
- diagnoseSkill
- grill-with-docsSkill
- install-merlin-skillsSkill
- merlin-skills-routingSkill
- playwright-cliSkill
- playwright-skillSkill
- gstack-qaSkill
- gstack-shipSkill
- speckit-analyzeSkill
- speckit-checklistSkill
- speckit-clarifySkill
- speckit-cliSkill
- speckit-constitutionSkill
- speckit-implementSkill
- speckit-planSkill
- speckit-specifySkill
- speckit-tasksSkill
- speckit-taskstoissuesSkill
- tddSkill
- to-prdSkill
- zoom-outSkill
The file
We split this AGENTS.md into the rules it gives an agent
9 rules. About 0 are common, 6 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Use the smallest skill chain that covers the work.
- Distinctive: line 23
When a feature uses spec-kit, the active feature directory is the source of truth:
- Very specific: line 39
Before starting Codex /goal:
- Very specific: line 43
Run create-goal to write specs/<feature>/GOAL.md.
- Very specific: line 44
Start /goal from that GOAL.md.
- Distinctive: line 46
Do not start long autonomous work from a loose chat summary when a spec directory exists.
- Distinctive: line 78
Before closing changes to this repo, run:
- Distinctive: line 84
If installer behavior changes, also run it against a temporary skill root:
- Distinctive: line 93
Do not claim the package is ready until validation passes.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Merlin Skills Agent Guide
This repository is a skill-first, /goal-first operating system for autonomous web coding. Treat it as a curated workflow layer, not as a grab bag of prompts.
Prime Directive
For any non-trivial task, load skills/merlin-skills-routing/SKILL.md before selecting implementation skills or creating a Codex /goal.
The router must start with a short Spec Kit consultation: inspect .specify/, active specs/<feature>/ artifacts, GOAL.md, task progress, and existing proof before choosing the next skill.
The default sequence is:
route -> Spec Kit consultation -> brainstorming when useful -> speckit-* artifacts -> create-goal -> /goal -> tdd -> gstack-qa -> browser proof -> gstack-shipIn this sequence, tdd is the implementation discipline for executing tasks.md slices after GOAL.md exists. It is not a replacement for speckit-tasks and it is not a bulk "write all tests first" phase.
Use the smallest skill chain that covers the work. Do not load every skill by default.
Source Of Truth
When a feature uses spec-kit, the active feature directory is the source of truth:
specs/<feature>/
spec.md
plan.md
tasks.md
GOAL.mdOptional source files may include research.md, data-model.md, contracts/, quickstart.md, and checklists/.
Generated goal files must reference the exact spec-kit artifacts they are based on.
Mandatory /goal Entry
Before starting Codex /goal:
- Run
merlin-skills-routing. - Confirm the feature has usable spec-kit artifacts.
- Run
create-goalto writespecs/<feature>/GOAL.md. - Start
/goalfrom thatGOAL.md.
Do not start long autonomous work from a loose chat summary when a spec directory exists.
Routing Table
| Need | Preferred skill route |
| Install or refresh Merlin Skills globally or for a project | install-merlin-skills |
| Install, verify, initialize, or manage GitHub Spec Kit CLI/integration | speckit-cli |
| Product idea, early concept, or "is this worth building?" | brainstorming then speckit-* |
| Ambiguous creative feature | brainstorming |
| Convert conversation to PRD | to-prd |
| Stress-test assumptions against docs | grill-with-docs |
| Understand a broad area | zoom-out |
| Create or refresh Spec Kit artifacts | speckit-constitution, speckit-specify, speckit-clarify, speckit-plan, speckit-tasks, then speckit-analyze/speckit-checklist when useful |
Build or fix with tests after tasks.md/GOAL.md exist, or for a small non-Spec-Kit fix with a clear behavior target | tdd |
| Debug hard bug or regression | diagnose |
| Create durable Codex goal | create-goal |
| QA and fix web app | gstack-qa |
| Browser proof or generated Playwright tests | playwright-cli |
| Custom Playwright automation | playwright-skill |
| Ship PR/release | gstack-ship |
Upstream Boundaries
Copied upstream skill sources should stay verbatim unless this repo explicitly owns the adaptation.
The intentionally adapted upstream ideas are create-goal, gstack-qa, and gstack-ship. Everything else should stay close to the selected upstream source or be excluded.
The full spec-kit repository is archived under vendor-snapshots/archives/ instead of extracted into the runtime skill tree. Only official generated Codex speckit-* skills are promoted into skills/; do not extract the whole upstream repository into the live tree.
Verification
Before closing changes to this repo, run:
npm testIf installer behavior changes, also run it against a temporary skill root:
SKILL_ROOT=/tmp/merlin-skills-root bash scripts/install-local.sh
find /tmp/merlin-skills-root -maxdepth 2 -name SKILL.md | sortGstack-derived install names must stay prefixed in the skill root. This package keeps only gstack-qa and gstack-ship; do not reintroduce the full gstack runtime surface.
Do not claim the package is ready until validation passes.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.