agentleFS
frostneyfrostney/knips/.agents/skills/deliver/SKILL.md

deliver

Skill1 starsChanged 11 days agoMITView on GitHub

Carries one feature, bug, issue, branch, or PR through verified delivery. Use when asked to deliver or ship a work item end to end, such as getting it merged or deployed, or when the user runs /deliver.

  • Commits and pushes

28 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

Used with

Projects with it also use prune, plan and 7 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • matthewpSkill

    prune

    Find and remove unused CSS in a Tideway project — classes defined in stylesheets but never referenced in markup. Use whenever the user wants to prune dead CSS, find unused styles or classes, clean up or slim a stylesheet, find orphaned BEM classes, or "tree-shake" hand-written CSS.

    Also in 3 projects that use this

  • BlockSkill

    plan

    Draft a TDD, PRD, or lightweight plan in the conversation without committing it.

    Also in 5 projects that use this

  • SentrySkill

    spec

    Produce a verified spec — problem, desired outcome, and acceptance criteria — before building. Use when given a GitHub/Linear issue link, when starting from a fuzzy in-conversation idea, or asked to "figure out what to do" / "scope this" / "what should I build" — whenever the *what* isn't yet pinned down. Grills the ambiguities, checks the premise against the code, and produces acceptance criteria you sign off on — then hands to design-first.

    Also in 4 projects that use this

  • AWSSkill

    ship

    Finalize dev work on an issue branch — update docs, push, and open a PR.

    Also in 5 projects that use this

  • AWSSkill

    wait-what

    Stop. That last message did not land: re-pitch it.

    Also in 3 projects that use this

  • Microsoft AzureSkill

    build

    Build the TypeSpec Rust emitter from source. Run this after cloning the repo, after pulling changes, or when TypeScript source files have been modified.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 3 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 28 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use /implement for development and required-behavior gaps.

    line 47
  2. Very specific:

    Use /create-pr for first publication and /update-pr for an existing PR.

    line 50
  3. Distinctive:

    Keep the same work item and reuse its open PR through repairs.

    line 51
  4. Very specific:

    Use /address-feedback to converge required reviews and current findings.

    line 57
  5. Distinctive:

    Use required companion skills when available.

    line 86
  6. Very specific:

    When only external transitions remain, such as CI, review automation or a deployment, park the work item: arm one wait for each awaited …

    line 112
  7. Distinctive:

    When the session goes on to another work item, start it from the recorded result and handoff, not from this item's working context.

    line 120

“Common” is measured: kept by at least 6 in 10 of the 28 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Deliver

Own one work item's outcome through its selected endpoint. Enter from its actual state, reuse settled decisions and valid evidence, and keep the parent task active when a child skill returns a fixable failure or pending transition.

Resolve scope and endpoint

Infer the repository and work item from the request, issue, bug description, confirmed idea, current branch or PR. Verify identities and existing work before mutation. Use /create-issue only when tracking is requested or required; an unfiled idea can go directly to /implement.

Honor an explicit endpoint: ready-to-merge, merged, or deployed. The default is deployed to the project's configured default integration destination, which may be production, staging, nightly or another established target. Read project instructions, delivery documentation and actual workflows to identify it; do not assume production or choose among ambiguous destinations.

A delivery request authorizes the in-scope development, validation, publication, feedback handling and Git operations needed for its endpoint, including merge and the configured integration path for deployed. Preserve explicit limits, repository protections and existing authorization. An earlier endpoint does not authorize later operations. Complete independent work before asking for a specific unresolved destination, provider, new infrastructure, spending or other material choice. Repair an existing integration workflow when an established in-scope defect blocks delivery; do not replace its provider or invent a new service as an incidental fix.

Coordinate the existing workflows

  • Use /implement for development and required-behavior gaps. It owns running and inspecting the result, code review, specification testing and in-scope repair. Pass the agreed outcome and fidelity criteria, not just a mechanism.
  • Use /create-pr for first publication and /update-pr for an existing PR. Keep the same work item and reuse its open PR through repairs. Reconcile uncertain writes before retrying; do not create a replacement PR merely because a call failed. If integration verification finds a required gap after merge, create a linked repair PR from the fresh default branch. Keep delivery active through its review, merge and integration verification; never rewrite merged history or try to update a closed PR.
  • Use /address-feedback to converge required reviews and current findings. Validate each finding against the agreed outcome; optional extras do not enlarge scope. Await pending checks and provider transitions through existing deterministic helpers. Inspect failures and route fixable causes through the development loop, then publish and verify the new exact head.
  • At ready-to-merge, require the complete readiness contract defined by address-feedback's readiness references for the selected PR's exact head: all verified requirement gaps resolved, current behavior and project evidence, successful required CI and reviews, and completed required thread handling. For a native stack, require its complete-stack readiness, including every needed fix layer. Ready for review alone does not satisfy this endpoint.
  • For merged or deployed, recheck readiness and use git-workflow to merge the authorized ordinary PR or complete native stack. Verify the resulting integrated revision. A partial prefix below a required fix layer is not done.
  • For deployed, run or await the existing integration path, then verify its destination, delivered revision or artifact provenance, and the required behavior in that environment. When later commits are included, verify that the delivered revision contains the integrated change and that its acceptance evidence still applies. A green deployment job or merge alone is insufficient.

Delegated work runs through the same skills. A worker packet routes each step it hands over through the skill that owns it, such as /implement, /create-pr, /update-pr or /address-feedback. It never prescribes raw publication or merge commands such as gh pr create or gh pr merge; those bypass the skill's gates, native stack detection and readiness checks. Each packet records the capability tier its lane runs on. Read capability tiers when you write a packet, run one, or receive a lane's escalation.

Use required companion skills when available; when one is unavailable, follow its reachable documented contract directly or report the specific missing capability. Do not invent a successful handoff. Reuse matching evidence for unchanged content, command, environment and requirements; rerun only missing or invalidated checks. Preserve independent review judgment.

Apply git-workflow's default-branch health gate before development starts and again before each merge. A default-branch repair it calls for is a separate work item that comes before this one.

Integration delivery and release publication are separate outcomes. Never invoke /create-release from this single-item loop or use a milestone release as a substitute for integration delivery. A configured nightly integration can satisfy deployed when its revision and behavior are verified. If the project has only a release path and no identifiable integration destination, finish independent work and ask for that concrete delivery decision.

Continue or finish

Continue through in-scope failures while a safe corrective action or useful investigation remains. When repeated attempts produce no new evidence or progress, reassess the cause and approach; escalate the specific remaining blocker after safe alternatives are exhausted. Do not loop unchanged probes, weaken acceptance or expand scope to manufacture progress.

When only external transitions remain, such as CI, review automation or a deployment, park the work item: arm one wait for each awaited transition under delivery-wait's running rules, record resumable state, and end the turn. Resume from each returned result, and arm the next wait before ending another turn. Never end a turn with a transition still expected and nothing armed to resume it. If the host cannot keep a wait armed, report the exact time or event to resume at.

When the session goes on to another work item, start it from the recorded result and handoff, not from this item's working context. Compact the context or start a fresh one at that boundary where the host allows it.

Finish only when the selected endpoint has observed evidence. An external blocker, unresolved material decision or needed authority can leave it incomplete; report the exact gap and next transition. Keep resumable state in the project's handoff convention. Return the work item, chosen endpoint, verified revision and destination when applicable, PR and integration evidence, and any remaining limitation. Match each completion claim to a returned result for that specific action and target. A successful merge does not confirm branch deletion: a deletion flag is a request, and deletion of an earlier branch says nothing about the repair branch. An acknowledgment likewise does not confirm cleanup or message delivery. Verify required outcomes; omit optional unconfirmed claims or label them unconfirmed, without implying failure. Apply this to the whole report, including tables and parenthetical remarks; see agent-writing for shared writing guidance.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.