agentleFS
Sign inStart free
ronniepinnellronniepinnell/casper/skills/backfill/SKILL.md

backfill

Skill1 starsChanged 4 months agoMITView on GitHub

Retroactively grade merged PRs against the claim-evidence discipline — which done-claims shipped with evidence, which shipped unproven. Zero-LLM (gh CLI); optionally seeds the verdicts ledger with UNVERIFIED BACKFILL rows. Use when adopting the toolkit on a repo with history, before trusting old "done"s, or to measure how the discipline is trending.

4 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the UNEVIDENCED list — those are the PRs whose "done" was asserted, never proven.

    line 30

“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/backfill — Grade the Dones You Already Shipped

The gates protect the next hundred PRs; history is where the false-dones you can already feel came from. This runs the claim-evidence check retroactively: every merged PR is graded EVIDENCED / UNEVIDENCED / NO-CLAIM, mechanically.

Invocation

/backfill                                  # last 100 merged PRs, current repo
/backfill --since 2026-01-01 --ledger .claude/verdicts.log   # window, not the whole history
/backfill --limit 300                                            # hard cap either way

Procedure

  1. Run python3 <toolkit>/scripts/backfill.py [--repo …] [--since DATE] [--limit N] — on a repo with thousands of PRs, ALWAYS pass --since (server-side filter); grade a window and widen only if the baseline needs it. (same claim words and evidence paths as the claim-evidence hook: an Evidence: line in the body, or test files in the diff).
  2. Read the UNEVIDENCED list — those are the PRs whose "done" was asserted, never proven. Spot-check the worst 3 against reality (the layer rule: observe the deployed behavior, not the diff).
  3. With --ledger, the summary + each UNEVIDENCED PR land as append-only BACKFILL | UNVERIFIED rows — history can be graded, not re-run, so retro rows are never CONFIRMED/REFUTED.
  4. The score is the baseline: re-run monthly; the trend is the report.
    BACKFILL: <repo> | <n> merged PRs | <c> done-claims | <e> evidenced (p%) | <u> shipped unproven

Rules

  • Report, never a gate — exit 0 always. The gate is the claim-evidence hook you install so the next PRs grade better.
  • Retro rows are UNVERIFIED by construction; upgrading one requires actually re-verifying the claim today (that's a /refute, logged separately).

Composes with

  • claim-evidence hook — the forward-looking half of the same check.
  • /refute — how an UNEVIDENCED PR's claim gets actually tested today.
  • /calibrate — the monthly trend of the backfill score.
  • /verdict — the ledger the --ledger flag writes to.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free