coagent8-workflow
Skill2 starsChanged 11 days agoMITView on GitHub
Using CoAgent8 and developing its task orchestration, integrations and release gates.
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- pr-mergeSkill
- coagent8-consultSkill
- coagent8-delegateSkill
- coagent8-doctorSkill
- coagent8-issueSkill
- coagent8-loopSkill
- coagent8-reviewSkill
- coagent8-sessionSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 30
Read GitHub Issues for the backlog and current state, ROADMAP.md for the direction and milestone order, and ARCHITECTURE.md for current …
- Distinctive: line 74
Keep author checks, independent review and owner approval distinct.
- Very specific: line 79
Update the owner in place: issues for remaining work/acceptance/evidence, roadmap for direction and order, architecture for …
- Distinctive: line 82
Keep required decisions in tracked documents.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
CoAgent8 workflow
Use the tool
For a simple question use consult with a self-contained proposal. Do not read this repository or run full diagnostics just to delegate arithmetic. Omit backend/model for ordinary defaults; select them when the user names one or supplies a constraint.
doctor lists models, efforts, output prices and approval per backend from cached data by default; scope health diagnoses sign-in and quota. A named model needs no backend: auto runs it where its CLI lists it. For a user's price limit pass output_price_below (USD per million output tokens, strict) and omit model to let the server choose; see the README's model selection guidance. On PRICE_LIMIT report the reason; never silently replace an explicit model. Current execution is read-only.
Let the server obtain model consent from the user. On unsupported/unanswered elicitation explain the model and reason. Do not set approval yourself or edit the governance.models lists to make a blocked call run. A default config path may differ from the running server's environment; do not change guessed config files.
Work on the project
Read GitHub Issues for the backlog and current state, ROADMAP.md for the direction and milestone order, and ARCHITECTURE.md for current contracts and target design. Read only the sections needed for the task. Optional docs/.local/ investigations are historical reference, never another plan.
The owner chose full agent orchestration on 2026-10-08. Roles and provider/model choices are independent; writing, bounded delegation and client skills require the roadmap's verified profiles/task contract. Do not remove current sandbox or nested-turn enforcement as a shortcut.
Follow core invariants, security, Git governance and the scope-based quality gate. Preserve the owner's local stack, existing changes, registrations and stored data. Audit blockers come before exposing affected capabilities. Work remains local unless the user authorizes integration/publication.
Branches and releases (rule 12): topic branches
start from develop and return to it by a squash-merged PR; there is no main.
Betas are tags on develop, release candidates and patches tags on an X.Y.x
line, which receives fixes only as backport PRs. You may prepare commits, a
release PR branch, release notes and gate evidence. Pushing, opening and merging a
PR follow the standing go-aheads in project.md;
tagging, GitHub Releases, npm and repository settings are the owner's.
Prose needs content/link review and git diff --check, not runtime tests or paid backend calls. Source comments/documentation tooling additionally need lint:docs and docs; executable changes need their scope's behavioral checks. Keep source, bundle, exact-artifact and native-host evidence distinct. A model READY string does not establish independent checks or release approval.
Prepare and maintain the PR description and checklist per rule 12 section 3, within the owner's authorization. Group related roadmap subtasks into one coherent result on one branch and PR, with its tests and documentation. Follow the roadmap's PR groups; task IDs and line counts do not automatically create PR boundaries. Provide the owner with current check evidence, the reviewed snapshot, unresolved findings and architectural decisions to assess. The owner's code review normally precedes opening the PR: prepare the local candidate and evidence first, address feedback, then open the PR when authorized. Carry actual pre-PR review evidence into the description without inventing approval or requesting the same owner review again for an unchanged snapshot. Keep author checks, independent review and owner approval distinct; a consult is advisory. GitHub Checks owns current CI status; do not duplicate it in a checklist. Use the PR template without copying rules or narrating the commit history.
Update the owner in place: issues for remaining work/acceptance/evidence, roadmap for direction and order, architecture for contracts/design. README serves users. Remove superseded reports rather than append transcripts. Keep required decisions in tracked documents; optional local notes may be ignored.
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.

