agentos-projects
Resolve and safely change an AgentOS project's local and remote lifecycle. Use before registering, cloning, creating, initializing, changing delivery or autonomy posture, removing, or reconciling a project checkout or provider repository.
- License 许可
- MIT
- 所属仓库
- akua-dev/agentos
- 最近更新
- 2026-11-09
- Stars
- ★ 8
- Forks
- 0
01 前言 Front Matter
- name
- agentos-projects
- description
- Resolve and safely change an AgentOS project's local and remote lifecycle. Use before registering, cloning, creating, initializing, changing delivery or autonomy posture, removing, or reconciling a project checkout or provider repository.
02 资源文件 1
- 参考文档agents/openai.yaml208 B
03 SKILL.md 正文
Manage an AgentOS project
Use PostgreSQL for the project registry and native Git plus the provider's reviewed CLI for repositories. Do not add an AgentOS project wrapper.
Resolve before mutation
- Resolve the source, unused destination, project scope, intended audience, reviewed delivery workflow, concrete review artifact, merge authority and autonomy posture. Record them as durable project prose, not an AgentOS mode enum. Agent-authored pull-request delivery defaults to no-mistakes when no exact reviewed project workflow selects another reviewed path. An absent, unknown or incomplete posture never silently falls back to a direct branch push and hand-opened pull request.
- Inspect existing registry rows, checkouts, remotes and relevant Second-Mate charters. Project access is non-exclusive; registration does not assign every task in a repository to one Mate.
- Ask before remote creation, credentials, cost, destructive removal or any outward effect not already covered by exact durable authority. An accepted remote-backed ship plus the recorded project workflow covers its task-branch commit, workflow-owned branch push and review-artifact creation or update; it never covers default-branch push or merge.
- Keep registry and checkout consistent. On partial failure, roll back only artifacts created by this operation and only when they contain no work.
Add or create
- For an existing repository, confirm the source and destination, clone with native Git into the selected persistent project home, verify its actual remote and instructions, then create the PostgreSQL project record.
- For local-only work, initialize an unused local path and create no unmentioned remote.
- Creating a GitHub repository requires explicit owner, name, visibility and
delivery approval. Inspect current
gh-axi --help, execute the approved operation synchronously and preserve its exit status and URL. - Do not encode Kun's
no-mistakes,direct-PRandlocal-onlynames as an AgentOS enum. Record the project's real reviewed workflow in durable prose; the default and exception above are workflow guidance, not a database mode. - For no-mistakes delivery, use the installed version's own Skill and live
no-mistakes axi --helpoutput. Verifyno-mistakes doctor, the actualorigin, the configured provider or fork push target and Assignment-scoped authentication before accepting the ship. If the checkout is not initialized, runno-mistakes initonly after those targets are resolved; never invent or create an unapproved remote. No-mistakes installs a local bare-repository Git proxy; it is not a hosted Git service. Its reviewed workflow may accept the local task-branch push, validate it, forward it to the configured provider remote and open the pull request.
Deliver and reconcile
Before dispatching remote-backed ship work, verify the actual Git remote,
selected workflow tools and provider authentication without printing or copying
credentials. If readiness is missing, keep the Task queued or blocked and load
$agentos-auth; do not create a worker whose Definition of Done is unreachable.
The selected project workflow owns validation, task-branch push, review artifact, local delivery and approval rigor. Risk can justify recommending a different path; it does not authorize a parallel Mate-added review gate. Use native Git, the selected workflow and reviewed provider commands directly, observe failure before updating local Task state, and retain full remote URLs. If readiness expires after implementation, preserve the committed branch and active Assignment while reporting the concrete blocker.
For a no-mistakes ship, choose a Crewmate setup that supplies this Skill and
name it in the brief. The same Crewmate that implements and commits the change owns
every no-mistakes axi run and no-mistakes axi respond call until the next
gate or terminal outcome. The native pipeline owns review, fixes, tests,
documentation, lint, rebase, task-branch push, pull-request creation and CI:
while it is active, the Crewmate never edits, commits, restarts or runs a
parallel review around it. Follow each response's live help instead of copying
version-specific mechanics here. Do not use --yes; route every ask-user
finding to the owning Mate through durable Inbox and continue only with the
exact returned decision. Report review readiness only at checks-passed, with
the full pull-request URL and the pipeline's fixes and risk. Merge authority
remains separate.
Remove
- Obtain explicit Captain authority.
- Inspect active and queued Tasks, Assignments, Second-Mate scope, worktrees, dirty state, unpushed commits, linked external work and other durable dependencies.
- Refuse removal while any dependency or unlanded work exists. A failed remote lookup is not positive proof that deletion is safe.
- Use only a reviewed guarded native path. If the selected release has none, report the capability boundary and preserve the registry, checkout and work.
04 同仓库更多 Skills
example-replacement
Reconcile the example replacement AgentOS distribution.
example-additive
Inspect the example additive AgentOS behavior.
agentos-upgrade
Upgrade one persistent AgentOS Mate or, from First Mate with exact Captain authority, every active persistent Mate in one Fleet to an exact published stable release, including its run-once released database migrations. Use for stable First- or Second-Mate updates, resumed post-replacement verification, and authorized rollback boundaries. Do not use for dogfood, Crewmates, provisioning, retired Mates, services, or another Fleet.
agentos-supervision
Reconcile and supervise an AgentOS Mate's direct reports across PostgreSQL, Kubernetes, Herdr, PVCs, and Git. Use at every First- or Second-Mate session start, after restart or compaction, while delegated work is active, when Inbox or runtime state changes, and for blocked, stale, stuck, failed, interrupted, or recovering agents.
agentos-secrets
Safely create, retry, rotate, take over, roll back, project, and revoke AgentOS-managed Kubernetes Secrets without retaining or printing credential bytes. Use whenever an AgentOS workflow moves credential files into Kubernetes, changes an existing managed Secret, verifies a projected credential, encounters conflicting Secret ownership or legacy annotations, or retires a Kubernetes-delivered credential.
agentos-runtime
Inspect, create, and operate AgentOS runtime pods, Kustomize overlays, Herdr sessions, Mise toolchains, agent worktrees, terminals, PVCs, Kubernetes health, attach paths and runtime recovery. Use for normal First- or Second-Mate Kubernetes operations, child-Agent workload creation, tool resolution, stuck-agent debugging, remote terminal access, pod or session recovery, and readiness diagnosis.