Coordinate substantial work with multiple outcomes, dependencies, handoffs, or durable checkpoints. Use when decomposing work, resuming a tracked effort, or revising a plan after evidence changes its premises. Reuse the active environment's work record and authorization.
83%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Passed
No findings from the security scan
The canonical home for this skill is planning-work in slurpyb/pi-taskfile
Turn requests into scoped, verifiable outcomes and carry them through the stage requested by the user. Deepen an existing plan only where coordination, risk, or resumption needs it; a clear small task can stay with the session baseline. Keep concise decision summaries, assumptions, and evidence in the existing record.
The active environment supplies tools, recording locations, ownership rules, approvals, and lifecycle transitions. Apply its instructions throughout. Use existing authorized capabilities and obtain specific approval for additional access, external effects, or expanded scope.
Read the request, applicable guidance, relevant artifacts, and current work record. Establish the requested stage: answer, research, review, plan, implement, or verify. Match actions and the deliverable to that stage.
Research discoverable facts first. Ask a focused question when the answer changes scope, success criteria, a consequential design choice, or authority to act. For reversible details within scope, state a reasonable assumption and continue.
Ready when: the outcome, requested stage, constraints, and consequential open decisions are explicit.
Account for four elements in each work unit's existing record or working context; use explicit fields when they make a handoff clearer:
For a quick answer, keep the compact contract in working context and respond directly; this specialist need not be loaded for that case. For work that needs tracking or resumption, record it through the active environment's supported interface. Reuse its authoritative work record. When durable recording requires an additional capability, provide the proposed record and a response checkpoint, and identify the persistence gap.
| Depth | Select when | Add to the compact contract |
|---|---|---|
| Compact | One clear outcome with established approach and checks | The four fields above |
| Structured | Work has distinct deliverables, dependencies, or handoffs | Context, work units, prerequisites, acceptance-to-evidence mapping |
| Design-intensive | Material alternatives, consequential uncertainty, broad impact, or costly recovery affect the approach | Options, decision rationale, risks, approval points, recovery |
Read task contracts when creating a structured or design-intensive plan, or when an existing record needs contract reconciliation. Read worked examples when choosing a boundary, depth, or output format needs a concrete demonstration; select the relevant case.
Ready when: every planned unit has a contract, and the selected depth covers the task's dependencies, decisions, and risks.
Keep each unit centered on one coherent outcome, including its checks and supporting documentation. Split when outcomes have distinct ownership, prerequisites, verification, or useful handoff boundaries. A planning step becomes a separately tracked item when independent accountability or resumption earns it.
For multiple units, read decomposition before setting order or proposing parallel work. Use actual prerequisite relationships and establish ownership of shared resources.
Ready when: units cover the requested outcome, prerequisites have a valid order, and each proposed handoff or parallel assignment has explicit boundaries.
For an execution request, proceed once the contract, prerequisites, ownership, and required approvals are ready. For other stages, produce that stage's artifact and apply its checks.
Choose evidence that observes the claimed outcome. Read verification before selecting unfamiliar checks, evaluating semantic outputs, handling a failed check, or reporting partial verification. Discover concrete commands and conventions from the environment.
Execute one bounded unit per owner at a time, check its result, and record the observed evidence before dependent work. Capture discovered work as a proposed follow-up. When a check fails, use its evidence to diagnose and correct the unit; when attempts repeat the same failure, revisit the premise. Pause the affected work when progress requires a new decision, capability, or authorization.
Ready when: each claimed result has appropriate observed evidence, and each remaining gap has an accurate disposition and next action.
Update the authoritative record before interruption, handoff, or context compaction. On resume, inspect live artifacts, ownership, approvals, and work state; reconcile them with the checkpoint. Recheck evidence affected by changes.
When evidence changes the plan, record the reason and affected units. Reconfirm approval for changes that expand scope, risk, access, or external effects.
For tracked work, provide a concise checkpoint or final handoff containing:
Report the requested stage complete when its completion condition is supported by evidence. Route acceptance, closure, publication, and deployment through the active environment's lifecycle. Finish with a proportionate response; for a compact answer, the answer and its relevant basis can carry the result directly.
Select the current decision; these are not additional mandatory phases.
slurpyb/pi-taskfile
808cef1
Canonical home
since Oct 27, 2026
Also appears in
since Oct 27, 2026
since Oct 27, 2026
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.