SkillCost Meter

SKILL.md cost estimator

SKILL.md Cost Estimator for Agent Skill Packs

A SKILL.md cost estimator reads the structure of a skill definition and predicts operational cost before the skill is executed. It looks for step count, tool breadth, credential dependency, context length, network behavior, and test coverage. This helps builders fix expensive design choices while the skill is still easy to edit.

Open the forecaster

When this matters

  • A Codex or Claude skill author wants a pre-publish check before sharing a skill pack.
  • A team reviews third-party skills and needs to know whether they are too broad or under-tested.
  • A marketplace wants contributors to pass a cost and QA gate before listing.

How to run the workflow

  1. Paste or upload SKILL.md and select the expected agent runtime.
  2. Detect steps, tool categories, file access, network calls, credentials, and approval language.
  3. Estimate context tokens from instructions, examples, references, and expected task payloads.
  4. Flag missing acceptance tests, broad permissions, and ambiguous completion criteria.
  5. Generate a revised cost forecast and pricing suggestion after edits.

Common risks

  • Static analysis cannot prove real runtime behavior unless it is paired with simulation.
  • Hidden dependencies in referenced scripts can change the cost and review burden.
  • Credentials and network access need explicit owner approval and auditability.

Where SkillCost Meter fits

SkillCost Meter starts with static SKILL.md analysis, then upgrades the same skill into a 20-run runtime forecast and red-line QA report.