AgentStack
SKILL verified MIT Self-run

Github Issues Kanban

skill-infolog-io-skills-github-issues-kanban · by infolog-io

>

No reviews yet
0 installs
9 views
0.0% view→install

Install

$ agentstack add skill-infolog-io-skills-github-issues-kanban

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Github Issues Kanban? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

github-issues-kanban

Purpose

GitHub Issues becomes the single source of truth for personal task tracking. Agents (you, subagents, other autonomous workers) can claim issues, do the work, and report back. The skill defines the protocol; the host implements dispatch.

Three concurrency primitives are first-class:

  1. Access lock — optimistic-concurrency claim with TTL (see references/lock-protocol.md)
  2. Dependency chain — DAG via depends-on:#N labels (see references/dependency-chain.md)
  3. Event bus — structured issue comments as the event log (see references/event-bus.md)

Profile

Full-shape skill. Uses spec references/ and assets/; uses marketplace conventions prompts/ and fixtures/. The protocol is agent-agnostic. Claude Code's Task tool appears only as one host implementation in dedicated sections.

Operating modes

| Mode | Trigger | Output | |---|---|---| | Audit | "audit my board", "/audit" | Scored board-health report; verdict + findings | | Triage | "triage my inbox", "/triage" | Conversational intake assigning labels + acceptance criteria | | Generate | "set up a board for X", "/generate" | New Project V2 from one of four archetypes | | List | "list my boards", "/list" | Multi-board summary with item counts | | Dispatch | "dispatch next", "what should I work on" | Picks next claimable issue per priority/age/size | | Claim (worker primitive) | "claim #N for agent X" | Acquires the lock atomically with optimistic concurrency | | Report (worker primitive) | "report result on #N" | Posts result event; transitions issue to ready-for-review |

Three concurrency primitives

Lock (claim with TTL)

status:claimed + claimed-by: + claim-expires:

Default TTL 30 min. Configurable per-issue via claim-ttl: label. Stale claims auto-release on next dispatch cycle.

Dependency chain (DAG)

depends-on:#  (multiple labels = AND)

Conductor refuses to dispatch issues whose deps aren't status:done. Cycles are detected and rejected with agent-skip label + error event.

Event bus (structured comments)

Event types: claimed, released, progress, blocked, result, error, yolo-dispatch, yolo-triage, yolo-disabled, stale-release. Schemas in assets/event-schema.json.

Confirm-first by default; YOLO bypass

Default: every dispatch surfaces a confirm prompt. YOLO mode bypasses confirms (per-dispatch, board-level via yolo:enabled label, or session-level). YOLO requires an audit-trail event per dispatch.

Auto-disable triggers: two consecutive blocked events, any error event during a YOLO dispatch, explicit user override, or session end.

See references/yolo-mode.md.

Four archetype templates

| Archetype | Columns | Use | |---|---|---| | personal-todo | Inbox → Next → Doing → Waiting → Done | Generic GTD | | writing-pipeline | Drafts → Research → Outline → Drafting → Review → Published | Long-form writing | | oss-triage | New → Triaged → Reproducing → Fix → Review → Closed | Bug + feature work | | sprint-planning | Backlog → Sprint → In-progress → Review → Done (WIP limits) | Time-boxed sprints |

Each archetype defines columns, WIP limits, default labels, and automation rules. See assets/template-*.json.

Verdict gates

| Verdict | Meaning | |---|---| | board-healthy | All audit dims ≥4; safe to enable YOLO; dispatch confidently | | drifting | One or more dims at 2-3; clean up before YOLO | | broken | Any dim at 1 or three at 2; do not dispatch until reconciled |

Triggers

| Phrase | Mode | |---|---| | kanban, board, /github-issues-kanban | (auto-route to mode) | | audit board, /audit, is my board healthy | Audit | | triage inbox, /triage, sort my new issues | Triage | | generate board, set up a board for X | Generate | | list my boards, /list-boards | List | | dispatch next, what should I work on, pick up work | Dispatch | | claim #N, /claim | Claim primitive | | report result on #N, /report | Report primitive | | yolo, enable yolo | Toggle YOLO mode (with confirmation) |

References

  • references/issue-as-task-contract.md — label scheme, status lifecycle, body requirements
  • references/lock-protocol.md — claim sequence, optimistic concurrency, TTL, release
  • references/dependency-chain.md — depends-on labels, transitive resolution, cycle rejection
  • references/event-bus.md — comment marker format, event types, payloads
  • references/conductor-protocol.md — dispatch algorithm, agent-agnostic with Claude Code section
  • references/worker-protocol.md — claim/work/report/release, agent-agnostic
  • references/yolo-mode.md — when to use, audit-trail requirements
  • references/audit-rubric.md — 6-dimension scored rubric

Prompts

  • prompts/audit-board.md — scored board health audit
  • prompts/triage-inbox.md — conversational intake for new issues
  • prompts/generate-board.md — create a new Projects V2 from archetype
  • prompts/list-boards.md — multi-board summary
  • prompts/dispatch-next.md — conductor primitive
  • prompts/claim-issue.md — worker primitive
  • prompts/report-result.md — worker primitive

Assets

  • assets/label-scheme.json — canonical label schema (status, claim, dependency, agent-output, sizing, priority, control)
  • assets/event-schema.json — event payload schemas per type
  • assets/audit-report-schema.json — structured audit output schema
  • assets/template-personal-todo.json — Inbox/Next/Doing/Waiting/Done
  • assets/template-writing-pipeline.json — Drafts→Research→Outline→Drafting→Review→Published
  • assets/template-oss-triage.json — New→Triaged→Reproducing→Fix→Review→Closed
  • assets/template-sprint-planning.json — Backlog→Sprint→In-progress→Review→Done

Interfaces

| Layer | Convention | |---|---| | Authentication | gh CLI (already auth'd by the user); host can swap to a token if needed | | State | GitHub Issues labels (current state) + comments (event log) | | Lock primitive | Optimistic concurrency via labels with TTL; hosts can plug stronger primitives | | Dispatch primitive | Host-specific (e.g., Claude Code's Task tool); the skill defines the protocol, not the mechanism | | Output formats | Markdown + JSON validating against schemas in assets/ |

Scope (v0.1.0)

  • All four archetypes
  • Parallel dispatch (up to host-configured max concurrency)
  • Optimistic concurrency with conflict resolution
  • TTL-based stale-claim recovery
  • Comment-based event bus (polling, not webhooks)
  • YOLO mode with audit trail and auto-disable

Deferred to v0.2+

  • True atomic lock via external service
  • Webhook-driven event bus
  • Cross-org board management
  • Team analytics
  • Direct GraphQL fallback (uses gh CLI exclusively)
  • Bulk operations
  • Custom field types beyond labels and status

Self-application

This skill must pass the canonical skill-structure audit at ≥4 on every dimension. The protocol's correctness is the load-bearing part — if the lock/dep/event-bus references aren't internally consistent, the audit penalizes accordingly.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.