agentsclimarketplace

Metrics

Skill MiaoY0uShan/FP/fp/metrics

Finish with Proof: a portable execution protocol for coding agents—diagnose before patching, bound delegation, and verify real outcomes.

Install
npx -y skills add MiaoY0uShan/FP --skill metrics

Assembled 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

Use only when routed by FP after a canonical Evidence Ledger and when measurement or a comparable baseline matters.

SKILL.md

2.3 KB, as published. Nobody here has run it

FP: Metrics

Metrics expose missing evidence; they do not manufacture proof that FP is efficient.

Use When

  • A canonical Evidence Ledger exists.
  • The user asks for measurement, a release needs evaluation, or adaptive improvement needs quantitative support.
  • Planned files and required checks can be obtained from the Execution Brief.

Skip trivial work. Without a required-check contract, verified task count is unknown. Without verified progress, TVP is undefined. Without a comparable baseline, make no improvement or efficiency claim.

Metrics

TVP = exact_context_tokens / verified_tasks_completed
Context Load Proxy = files_read + observed_skills_loaded + observed_reports_used
Proxy TVP = context_load_proxy / verified_tasks_completed
Scope Creep Rate = unplanned_files_touched / files_touched
Verification Rate = tasks whose required checks all passed / completed tasks
Rework Rate = failed_or_reopened_tasks / completed tasks

Counts must come from metrics_inputs, the brief, or ledger arrays. Missing values stay unknown/null; never hardcode defaults.

For one task, verified progress requires all of:

  • result=pass;
  • a brief with at least one required check;
  • every required check was run and passed;
  • no failed check;
  • at least one evidenced verified claim;
  • no contract, scope, or context-budget violation.

Evaluation Quality

When comparing approaches, hold task, agent/model, repository revision, and acceptance checks constant; use fresh isolated workspaces, repeat runs, preserve raw artifacts, and score functional verification/safety separately from tokens or lines changed. Disclose contamination, timeouts, sample size, and limitations.

Procedure

  1. Validate the canonical ledger and brief.
  2. Read observed metric inputs; do not estimate missing counts.
  3. Calculate only defined metrics.
  4. State facts separately from comparisons.
  5. Recommend adaptive improvement only when evidence supports a reusable change.

Run:

node fp/metrics/collect.js <evidence-ledger.json> [execution-brief.json]

Use templates/metrics-report.md for a manual report. The collector supports legacy Markdown for one migration cycle and prints a warning; JSON v1 is canonical.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.