warchief
Skill1 starsChanged 52 days agoApache-2.0View on GitHub
Have the AI choose the next orchraft lifecycle step and hand off to the right skill, without performing git or GitHub mutations. Use when the user says warchief, asks what should happen next, or wants orchraft to route the work.
The file
Show its rules and text
We split this skill into the rules it gives an agent
25 rules. About 0 are common, 7 are distinctive, and 18 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
18
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Use this skill when the user wants the clan to decide what should happen next across the software lifecycle, instead of naming a specific …
- Distinctive: line 13
Read the user's goal and the current repo state.
- Distinctive: line 19
Stop before every git, PR, issue, or release mutation unless the current message explicitly approves that exact mutating workflow, or the …
- Distinctive: line 29
Treat user-facing role names as portable workflow roles.
- Very specific: line 34
Use Forgehand for implementation and tests.
- Very specific: line 36
Use Warpath when the user explicitly asks to repeat fix → ship → roast until the latest roast reports zero findings.
- Very specific: line 39
Use Trialmaster for read-only review findings.
- Very specific: line 41
Use Tactician / warplan when the work needs a grounded implementation plan before code changes.
- Very specific: line 43
Use Quest-Giver / quest when the next useful artifact is an issue or issue update.
- Very specific: line 45
Use Raid Captain / ship only when the user explicitly asks to branch, commit, push, or open/update a PR in the current turn.
- Very specific: line 47
Use Trialmaster / roast when the user asks for PR review or the PR is ready for review.
- Very specific: line 49
Use Rune-Reader / runes when the user asks for status, resume, or where work stands.
- Very specific: line 51
Use Scout / yap when the user asks to explain a file, diff, PR, policy, or error.
- Very specific: line 53
Use Loremaster / lore when the user asks to add or clean up comments/docstrings.
- Very specific: line 55
Use Chronicler / chronicle when the user asks to write or update a standalone Markdown document.
- Very specific: line 57
Use Haulmaster / land only when the user explicitly asks to merge/land in the current turn.
- Very specific: line 59
Use Herald / herald when the user asks for release notes or what a release shipped.
- Very specific: line 61
Use Reckoner / reckoning when the user asks which review findings were declined rather than fixed.
- Distinctive: line 66
If more than one route fits, choose the earliest lifecycle step that removes real uncertainty.
- Very specific: line 72
Read context/policies/approval-policy.md before any route that could mutate git, PRs, issues, or releases.
- Very specific: line 74
Never treat the warchief request itself as approval to commit, push, open/update/merge a PR, resolve a thread, mutate an issue, or write a …
- Very specific: line 77
Do not create a stored state file.
- Very specific: line 78
Do not chain multiple mutating skills from one approval.
- Distinctive: line 92
If that role is mutating and neither the current turn nor a verified Autonomous Mode entry authorizes it, ask for the exact go-ahead …
- Distinctive: line 95
If a non-mutating role fits and no approval is needed, run it directly.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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
Warchief Workflow
Use this skill when the user wants the clan to decide what should happen next across the software lifecycle, instead of naming a specific skill.
At A Glance
- Read the user's goal and the current repo state.
- Choose the next fitting orchraft role:
Quest-Giver,Tactician,Forgehand,Warpath,Raid Captain,Trialmaster,Haulmaster,Herald,Rune-Reader,Reckoner,Scout,Loremaster, orChronicler. - Explain the handoff briefly, then run only the step the approval policy says is authorized.
- Stop before every git, PR, issue, or release mutation unless the current message explicitly approves that exact mutating workflow, or the verified Autonomous Mode from the applicable approval policy or active config names that exact action. For configs, apply the shared verification rules, including the trusted exact-path ignore check for a local config absent from the default branch. A failed or mismatched verification is not authorization.
Routing
- Treat user-facing role names as portable workflow roles. Run the matching
skill on every host; when the host exposes a compatible named subagent,
delegate to its adapter:
forgehand,tactician,scout,loremaster,trialmaster,chronicler, orrune-reader. These agents stop before mutations that require separate approval. - Use
Forgehandfor implementation and tests. It may edit files but never commits, pushes, opens a PR, posts a review, merges, or changes an issue. - Use
Warpathwhen the user explicitly asks to repeat fix → ship → roast until the latest roast reports zero findings. It re-reads live state after each cycle, pauses at approval gates, and hands off a clean PR toland. - Use
Trialmasterfor read-only review findings. It may inspect code and useroast's policy, but never posts the review. - Use
Tactician/warplanwhen the work needs a grounded implementation plan before code changes. - Use
Quest-Giver/questwhen the next useful artifact is an issue or issue update. - Use
Raid Captain/shiponly when the user explicitly asks to branch, commit, push, or open/update a PR in the current turn. - Use
Trialmaster/roastwhen the user asks for PR review or the PR is ready for review. - Use
Rune-Reader/runeswhen the user asks for status, resume, or where work stands. - Use
Scout/yapwhen the user asks to explain a file, diff, PR, policy, or error. - Use
Loremaster/lorewhen the user asks to add or clean up comments/docstrings. - Use
Chronicler/chroniclewhen the user asks to write or update a standalone Markdown document. - Use
Haulmaster/landonly when the user explicitly asks to merge/land in the current turn. - Use
Herald/heraldwhen the user asks for release notes or what a release shipped. - Use
Reckoner/reckoningwhen the user asks which review findings were declined rather than fixed.
Direct skill names remain supported alongside these role aliases.
If more than one route fits, choose the earliest lifecycle step that removes real uncertainty. For example, plan before editing, status before shipping an unknown branch, and review before landing.
Guardrails
- Read
context/policies/approval-policy.mdbefore any route that could mutate git, PRs, issues, or releases. - Never treat the
warchiefrequest itself as approval to commit, push, open/update/merge a PR, resolve a thread, mutate an issue, or write a release. - Do not create a stored state file. Use
runesfor live status. - Do not chain multiple mutating skills from one approval. A clean
roastmay makelandthe next recommendation, butlandstill needs its own current-turn go-ahead. Warpathis the explicit exception for orchestration, not for approval: it may coordinate repeated lifecycle steps, but each mutation remains gated by the approval policy.
Handoff
- Unless the user asked for plain output, open with a fresh one-line phrase
in the orc voice, speaking as the Warchief from
context/personality.md(the target repo's copy when it exists, otherwise${CLAUDE_PLUGIN_ROOT}/context/personality.md). - Name the selected next role and why.
- If that role is mutating and neither the current turn nor a verified Autonomous Mode entry authorizes it, ask for the exact go-ahead instead of running it.
- If a non-mutating role fits and no approval is needed, run it directly.
Examples
Chief, the council calls for `runes` first: this branch's PR state is unknown.
Reading the board before we swing.Chief, `ship` is the next march, but that means branch/commit/push/PR work.
Say "ship it" and we move.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.

