Pm tasks core
Collection of Agent Skills for Claude Code, Cursor, Codex, Windsurf, and any agent that speaks the open Skills spec. Ships via npm + plugin + skills add.
npx -y skills add llodev/skills --skill pm-tasks-coreAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Core extraction + vocabulary for the @llodev/pm-tasks-* family. Use when working with any pm-tasks-<tool> adapter (Trello, Asana, etc.) — provides Phases 1–3 (identify input, extract structure, build the generic card) plus the canonical CRUD vocabulary (task.create, checklist.check, task.close, task.due-date.set, task.assignee.add, task.comment.add, task.move, task.parent.set, task.estimate.set, task.sprint.set) consumed by adapters. Also defines autonomous-mode contract (sentinels, allowlist, scope, audit log) and the shared init UX. Triggered indirectly by any prompt that an adapter handles (e.g. "create Trello card", "publish plan to Asana", "[autonomous] create task"). Do NOT activate alone — it has no tool-specific formatting.
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
4.7 KB, as published. Nobody here has run it
pm-tasks-core
Shared core for all pm-tasks-<tool> adapters. Defines the extraction phases, the generic-card structure, the CRUD vocabulary, the autonomous-mode contract, the configuration lookup rules, and the audit-log format. Adapters reference this skill by path — no formal dependency mechanism in the spec.
Routing
Adapters invoke this skill BEFORE applying their tool-specific formatting. The exact pointer is documented in references/contract.md.
Phases
| Phase | Purpose | Reference |
|---|---|---|
| 1 | Identify the input (plan file vs inline paste vs implicit) | references/contract.md § 1 |
| 2 | Extract sections by intent (goal, prereqs, tasks, done-when) | references/contract.md § 2 |
| 2.5 | Anti-patterns gate | anti-patterns/core.md |
| 3 | Build the generic card | references/generic-card.md |
Adapters then execute Phases 4+ per their own SKILL.md.
CRUD vocabulary (verbs adapters implement)
See references/crud-vocabulary.md. Ten verbs: task.create, checklist.check, task.close, task.due-date.set, task.assignee.add, task.comment.add, task.move, task.parent.set (adapter-optional), task.estimate.set (adapter-optional), task.sprint.set (adapter-optional; Linear first supporter). All idempotent (with clientToken rules for non-natural cases).
Autonomous mode
See references/autonomous-mode.md. Activated only by sentinel [autonomous] / --auto / env LLODEV_PM_TASKS_AUTONOMOUS=1. Requires explicit allowlist + scope + rate limit in the tool's config. Never inferred. Agent-authored copy (commits, PRs, comments) follows the workspace narration language — see Narration language.
Narration language
Narration language. When .<tool>.json sets locale, all agent-authored narration — comments the agent writes, autonomous-mode commit and PR copy, code comments, and audit/CLI free-text — MUST be in that locale's language. The plan's language still governs task content (title, description, checklist items); these are separate concerns. If locale is unset, pm-tasks-core-doctor (C-LANG-1) warns and narration defaults to English.
Configuration
Lookup order: <git-root>/.<tool>.json → ~/.config/llodev/pm-tasks/<tool>.json → abort. Secrets NEVER in JSON (env vars / OS keychain only).
Audit log
Append-only JSONL at ~/.local/share/llodev/pm-tasks/<tool>/audit.log. Schema in references/audit-log-format.md. Doubles as the lookup index for <task-ref> resolution.
Init helper
Adapters expose npx @llodev/pm-tasks-<tool> init. Shared UX in references/init-ux.md. Implementation library at @llodev/pm-tasks-core/init-lib (TypeScript source under ./src/init-lib.ts; compiled to ./dist/init-lib.js).
Health check (doctor)
pm-tasks-core-doctor validates a workspace config, autonomous allowlist, audit writability, and network reach. Check matrix and exit codes in references/doctor.md — pm-tasks-core-doctor check catalog + exit codes.
Standalone fallback
This skill is not useful without an adapter. If activated alone, tell the user to install at least one pm-tasks-<tool> package.