simiancraft/simiancraft-skills / skills/burn-down-github-issues
burn-down-github-issues
UnverifiedRun an unattended burndown of a GitHub issue backlog as cards on a per-operator GitHub Projects board whose lanes are an explicit state machine: appraise and size the inbox, carve what is over the ceiling, fix the ready ones in parallel git worktrees, prove each fix on a draft pull request, have an isolated second-engine reviewer judge it, land one branch at a time with freshness checks, and dead-letter per phase whatever a machine cannot finish. Use when the task is "work the backlog unattended", "triage and fix recent issues", "run the issue loop", "burn down issues", or "where is issue N on the board". Requires a per-repository burn-down-github-issues.config.ts at the target repo's root (the loop refuses to start without it), the prove-work-on-github skill, the codex and claude CLIs for the default seats, and a gh token with the project scope for the board. Skip for one-off fixes to a known issue you are already working interactively, and for backlogs whose issues are mostly product decisions rather than code.
lib/admission.ts
View on GitHub/**
* Mid-run admission: the work phase's refill. A run's work list used to be fixed when the run
* started, so anything the run itself made eligible (a dependent whose blocker it merged, a child
* the knife carved, a trunk's released remainder) waited for the next run. After a lane settles,
* and once more before the run calls itself idle, the backlog is placed again from the tracker and
* what is newly Ready joins the run. Placement, holds, claims, blockers, and the size ceiling decide
* exactly as they do at the start; this module only decides what is new and what the budgets allow.
*/
/** The slice of a placement this module reads: the lane each issue is in, and why. */
export type PlacedIssue = {
number: number;
blockedBy?: { nodes: Array<{ number: number; state: string; stateReason?: string | null }> } | null;
};
export type AdmissionPlacement<I extends PlacedIssue> = Map<number, { lane: string; why: string; issue: I }>;
export type AdmissionDeps<I extends PlacedIssue> = {
/** `--limit`: every issue handed to a worker this run counts, the first batch included. */
limit: number;
/** `--appraise-limit`: every issue appraised this run counts. */
appraiseLimit: number;
/** False under `--no-appraise`, a dry run, or when the start of the run could not size. */
canAppraise: () => boolean;
stopping: () => boolean;
/** Fetch the base, so appraisers judge, and lanes branch from, what the last merge left. */
fetchBase: () => void;
/** Place the backlog from the tracker (the run's `placeBacklog(allIssues())`). */
place: () => Promise<AdmissionPlacement<I>>;
/**
* Appraise the inbox lane, leaving out `skip`, at most `limit` issues; the numbers it appraised,
* or null when the appraisal lock is held elsewhere and this pass must not size anything.
*/
appraise: (placement: AdmissionPlacement<I>, skip: (issue: I) => boolean, limit: number) => Promise<number[] | null>;
/** The run's selection over Ready, in its order, without logging every exclusion again. */
select: (placement: AdmissionPlacement<I>) => I[];
log: (line: string) => void;
};
export type Admission<I extends PlacedIssue> = {
/** Issues handed to a worker this run. */
readonly dispatched: Set<number>;
/** Issues appraised this run. */
readonly appraised: Set<number>;
/** Records the start-of-run batch, so the refill neither repeats it nor forgets it in the budget. */
start: (issues: Iterable<number>) => void;
/** The refill: newly eligible issues, within the budgets, each admitted at most once per run. */
admit: () => Promise<I[]>;
};
/** Rounds of sizing one refill makes: freed cards, then children their carving created, and so on. */
export const MAX_APPRAISAL_PASSES = 4;
export function createAdmission<I extends PlacedIssue>(deps: AdmissionDeps<I>): Admission<I> {
const dispatched = new Set<number>();
const appraised = new Set<number>();
const reportedStuck = new Set<number>();
/**
* A card waiting on a blocker closed without completing never leaves W1 by itself (that branch of
* `BLOCKER_CLOSED` is not built), so the run says so once instead of leaving it silent.
*/
const reportStuck = (placement: AdmissionPlacement<I>) => {
for (const { lane, issue } of placement.values()) {
if (lane !== 'W1' || reportedStuck.has(issue.number)) continue;
const blockers = issue.blockedBy?.nodes ?? [];
if (blockers.length === 0 || blockers.some((b) => b.state !== 'CLOSED')) continue;
const dead = blockers.filter((b) => b.stateReason !== 'COMPLETED').map((b) => `#${b.number}`);
if (dead.length === 0) continue;
reportedStuck.add(issue.number);
deps.log(`#${issue.number} waits on ${dead.join(', ')}, closed without completing; a person decides whether it still applies`);
}
};
return {
dispatched,
appraised,
start: (issues) => {
for (const n of issues) dispatched.add(n);
},
admit: async () => {
if (deps.stopping()) return [];
const room = deps.limit - dispatched.size;
if (room <= 0) return [];
deps.fetchBase();
let placement = await deps.place();
reportStuck(placement);
// Sizing can create work: an oversized card goes to the knife, whose children arrive unsized.
// Size again until nothing new is unsized, so a refill asked at idle does not end the run with
// freshly carved children nobody will appraise.
for (let pass = 0; pass < MAX_APPRAISAL_PASSES; pass++) {
const budget = deps.appraiseLimit - appraised.size;
const unsized = [...placement.values()].some(({ lane, issue }) => lane === 'A1' && !appraised.has(issue.number));
if (!unsized || budget <= 0 || !deps.canAppraise() || deps.stopping()) break;
const sized = await deps.appraise(placement, (issue) => appraised.has(issue.number), budget);
if (sized === null) {
deps.log('the appraisal lock is held elsewhere; cards freed mid-run are sized by the next run');
break;
}
for (const n of sized) appraised.add(n);
placement = await deps.place();
if (sized.length === 0) break;
}
if (deps.stopping()) return [];
const admitted = deps
.select(placement)
.filter((issue) => !dispatched.has(issue.number))
.slice(0, room);
for (const issue of admitted) {
dispatched.add(issue.number);
deps.log(`#${issue.number} admitted mid-run: ${placement.get(issue.number)?.why ?? 'Ready'}`);
}
return admitted;
},
};
}
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |