milestone
Skill9 starsChanged 2 days agoMITView on GitHub
Implement the pohunek milestone specified by a GitHub issue in a fresh worktree, delegating the build to subagents, then run the full gate set. Use when the user points at a GitHub issue and asks to implement that milestone.
10 other projects use a version of this.
Adopted by
Most recently taken up by
- emingo/claude-skillstheir own version, first seen Nov 16, 2026
- Jasenpan1987/multi-codetheir own version, first seen Nov 8, 2026
- ArcticFox2029/chamnantheir own version, first seen Nov 5, 2026
- omnigami-ai/milestonetheir own version, first seen Nov 1, 2026
- frbar/claude-skillstheir own version, first seen Oct 15, 2026
- enjunear/skillstheir own version, first seen Oct 13, 2026
Other versions10 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 10
- ArcticFox2029/chamnan★ 10same name
- Eipi-Tong/agentic-coding-starter-kit★ 3same name
- ahamedzoha/atelier★ 2same name
- j4rk0r/claude-skills★ 2same name
- gihwan-dev/claude-code-gui★ 1same name
- Jasenpan1987/multi-code★ 1same name
- emingo/claude-skills★ 0same name
- enjunear/skills★ 0same name
- frbar/claude-skills★ 0same name
- omnigami-ai/milestone★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository13 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- ms-rustSkill
- contributeSkill
- coordinateSkill
- deliver-issueSkill
- gatesSkill
- github-workflowSkill
- merge-advanceSkill
- milestone-reviewSkill
- plan-phaseSkill
- pr-handoffSkill
- releaseSkill
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 9 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
- Very specific: line 31
Create a worktree off
main: Use the script, which fetches origin/main, creates pohunek-worktrees/<milestone-slug> beside the primary … - Very specific: line 45
Read the Rust guidelines first — mandatory: Before creating or modifying ANY .rs file, read the applicable files from …
- Very specific: line 47
Use .agents/rust-guidelines/SKILL.md as the which-file-when index.
- Distinctive: line 68
Every slice must pass the gates on its own and be complete production code for its concern — no stubs, dead code, or placeholders a later …
- Very specific: line 93
Keep the assistant knowledge bundle current: If the milestone changes a CLI command/flag, a protocol method/event, an operating-model …
- Distinctive: line 117
No PoC, no minimal/partial versions, no shortcuts unless the user explicitly asks.
- Distinctive: line 119
No mocks or stubs for specified functionality.
“Common” is measured: kept by at least 6 in 10 of the 9 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.
Files
milestone — implement a milestone from its GitHub issue
Implements one milestone specified by a GitHub issue, following the repo's
standing loop: fresh worktree off main, guideline-compliant implementation
delegated to parallel workers, then the full gate set green before hand-off.
The issue is the spec — there is no NEXT.md.
Steps
- Resolve the issue. Route through the
github-workflowskill. Use the issue URL/number the user or task explicitly provides; when none is given, search for matching issues and reuse a unique open match. If none matches and the requested scope is concrete, automatically create the issue with its scope and DoD and add it to the configured project. Ask only when the scope or competing matches are genuinely ambiguous. Fetch the live issue body and comments — later comments may hold decisions, follow-ups, and evidence that refine the body. Extract the scope and the DoD items with their stable IDs — these are the testable success criteria. If they are ambiguous, resolve the ambiguity before implementation; do not invent scope. Core ships no UI (project constraints (.agents/instructions/project.md)): a milestone that needs a GUI, web, or launcher change is work forzajca/pohunek-work. Create a worktree off
main. Use the script, which fetchesorigin/main, createspohunek-worktrees/<milestone-slug>beside the primary checkout (from any checkout; it prints the absolute path) onzajca/<milestone-slug>, and reflink-seeds itstarget/from the main checkout so the first build compiles only workspace crates:scripts/worktree-new <milestone-slug>It fails closed when reflinks are unavailable or a build holds the main checkout's Cargo lock;
--no-seedaccepts a cold build instead. Never put a worktree under/tmp(a small RAM tmpfs). (Never implement directly onmain. Do not disturb other checkouts or unrelated active worktrees.)- Read the Rust guidelines first — mandatory. Before creating or modifying
ANY
.rsfile, read the applicable files from.agents/rust-guidelines/. Use.agents/rust-guidelines/SKILL.mdas the which-file-when index; at minimum11_universal_guidelines.md, adding02_application_*,03_correctness_*, and06/12/13/14/15(library design) as the task warrants. ApplyM-CANONICAL-DOCS, short names, documented magic values,#[expect(..., reason = "...")]over#[allow]. Update the// Rust guideline compliant <date>marker on any file you bring fully into compliance. Plan the PR stack. Split the milestone into ordered slices per
pullRequestsin.github/agent-workflow.json: one coherent concern per slice, in dependency order (e.g. protocol types → daemon → client/cli consumers), each mapped to the DoD items it satisfies. Record the slice plan in the issue body (via thegithub-workflowskill) before implementing. Build each slice on its own branch based on the previous slice's branch, in the same worktree:git switch -c zajca/<milestone-slug>-1-<concern> # off the step-2 branch # ...implement, gate, commit slice 1... git switch -c zajca/<milestone-slug>-2-<concern> # off slice 1Every slice must pass the gates on its own and be complete production code for its concern — no stubs, dead code, or placeholders a later slice fills in. A milestone that is already one small concern is a stack of one.
- Implement via parallel subagents by default. Decompose the
slice and delegate implementation to parallel subagents — this
is the default, not something to wait for permission on. Brief each worker
with concrete
path:linecontext (per the global briefing protocol); they start with a clean context window. If the wire protocol (crates/protocol) changes, expect ripples inclient,daemon, andcli— update and test all of them, plusdocs/public-api.md. - Test cooperating behavior, not functions (testing policy (
.agents/instructions/verification.md#testing-policy)). New observable behavior gets an integration scenario — multiple production components cooperating through the supported boundary its callers use — or an E2E scenario driving the real product processes; a bug, lifecycle, durability, or security fix gets a regression scenario at such a boundary; a trivial helper change or an already-covered refactor gets no new scenario. Writing a new unit test (an isolated component or per-helper scenario, whatever file it lives in) is prohibited. Extend the existing protocol/state-machine suites before creating new ones, and only with scenarios that themselves meet the integration/E2E boundary — appending to an existing suite does not exempt a test from the boundary, so never add unit-tier tests to those suites; no coverage targets. Iterate withcargo t; when several agents or worktrees share the host,cargo ta(cargo xtask affected) runs only the fast tests of the changed crates and their dependents and saves CPU. Neither replaces the gates in step 8. - Keep the assistant knowledge bundle current. If the milestone changes a
CLI command/flag, a protocol method/event, an operating-model
concept, a safety rule,
docs/public-api.md, or a path indocs/knowledge/assistant/source-map.md, update the matchingdocs/knowledge/file in the same change. A stale bundle is treated like stale code. - Run the gates. Invoke the
gatesskill once on the final revision of each slice; a fix re-runs only the checks whose inputs it changed. Iterate until every gate is green. Report honestly — never claim green without running it. - Record progress in the issue. Via the
github-workflowskill, post per-DoD-item results withpath:lineevidence and the gate results as issue comments as major steps complete; keep the project status consistent (In Progress while implementing). When the work stops mid-run (blocked, interrupted, or handed over), leave the handoff comment the skill specifies: branch/worktree, revision, scope covered vs remaining, the exact checks run with their real exit results (including skipped/failed ones and why), blockers, and any subagent/worker run IDs. - Report. Summarize what was implemented against each DoD item with
path:lineevidence per slice, and state the gate results. Do not push, open PRs, or merge here — that is thepr-handoff/merge-advancepath.
Constraints
- No PoC, no minimal/partial versions, no shortcuts unless the user explicitly asks. Implement the milestone's full scope.
- No mocks or stubs for specified functionality; if blocked, ask.
- Local unsigned commits on the slice branches are part of building the stack (the next slice branches off them). Pushing, opening PRs, and merging happen only when asked; this skill stops at "stack implemented + every slice gates green", recorded as evidence on the issue.
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.

