SuperAPI AI Directory

pc-estimation

General & other

View repository

Updated 10-16 06:45

About this skill

Estimation

Estimate to expose uncertainty and trade-offs, not to pretend the future is certain.

Context

Estimation converts a task list into a planning signal the team can actually use.

See context and anti-pattern notes.

Inputs

I/O contract notes define required inputs and authority.

Process

Step 1: Identify the Planning Decision

Estimate only enough to compare scope, sequence work, or assess a requested deadline. A single authorized short task may need only an uncertainty note unless the route requires more. Reuse accepted estimates when scope, dependencies, capacity, and execution assumptions still match.

Step 2: State Units and Evidence

Choose compatible units for the decision. Keep relative size separate from elapsed time; never add story points to hours or convert them without calibration. Use comparable completed work when available and state differences.

Separate active work, external waiting, review/integration, and risk contingency. For agent-assisted execution, identify human review and tool/runtime dependencies; do not convert human days into agent minutes or assume speedup without comparable measured runs.

Step 3: Size the Work With Visible Uncertainty

For each relevant task, record a range or size bucket, confidence, assumptions, and evidence. Mark unresolved work unknown when a credible range is unavailable; name the fact or bounded investigation that would make it estimable. Do not manufacture a point estimate to fill the table.

Include shared setup and dependencies once. Show sequencing constraints so parallel-looking tasks do not imply elapsed-time savings they cannot deliver.

Step 4: Hand Off the Planning Signal

Publish the estimate-set under its I/O contract. Distinguish schedulable from blocked or low-confidence work and state re-estimation triggers. The estimate informs pc-sprint-planning; it does not promise delivery or assign capacity.

Outputs

Produce only declared outputs at their documented quality boundary.

Quality Gate

  • Relevant tasks have comparable ranges/buckets or explicit unknowns
  • Units, evidence, confidence, dependencies, and assumptions are visible
  • Active work and waiting are distinguished without invented agent-speed claims
  • The consumer can identify schedulable work and re-estimation triggers

Distribution

  • Public install surface: skills/.curated
  • Canonical authoring source: skills/03-planning/pc-estimation/SKILL.md
  • This package is exported for npx skills add/update compatibility.
  • Packaging stability: beta
  • Capability readiness: beta
  • Portability: portable_with_caveat
  • Public caveat: Portable as skill guidance; full governance guarantees require the Prodcraft repository contracts and validation checks.