Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Get Started Free →

Observable Receipt Distiller · learnprompt | DecimalAI

agent toolingby learnprompt

Observable Receipt Distiller

learnprompt/observable-receipt-distiller

Source learnprompt/learnprompt ↗License NOASSERTION

Distill a candidate skill from observable repair receipts, then prove it on holdout fixtures.

NewMaintained

Automatic

AI Engineer pack

Last verified 8/13/2026

What it changes

7 of 24 test cases passed with the skill.

Pass-rate lift: —
Tokens: +75%
Agent turns: 0%
Response time: +67%

Model gemini-3.6-flashMeasured 8/13/2026v17 of 24 cases passed with skill · 5 comparable

Tokens, turns and time describe resource use, not a dollar price.

Lift withheld: 5 comparable cases; this report requires 8.

Use this skill

Save it into your project's skills folder. Claude Code picks it up in its next session.

curl -fsSL --create-dirs -o .claude/skills/learnprompt-observable-receipt-distiller/SKILL.md https://app.decimal.ai/s/learnprompt-observable-receipt-distiller/SKILL.md

Want it pinned and updatable? pip install "decimalai>=0.15", then decimalai skills add learnprompt-observable-receipt-distiller, and start Claude Code with claude --plugin-dir "$(git rev-parse --show-toplevel)/.agents"

1 · One-time — install the CLI (docs ↗)

pip install "decimalai>=0.15"

2 · Get the skill — writes .agents/skills/learnprompt-observable-receipt-distiller/ and pins it in .decimal/skills.lock

decimalai skills add learnprompt-observable-receipt-distiller

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/learnprompt-observable-receipt-distiller/SKILL.md https://app.decimal.ai/s/learnprompt-observable-receipt-distiller/SKILL.md

Raw file: /s/learnprompt-observable-receipt-distiller/SKILL.md ↗ · machine-readable summary at /s/learnprompt-observable-receipt-distiller.json ↗

Safety checks: 3 of 3 passed

Code scan CleanIntent PassedContent Clean

See it in action

One of 4 test cases the skill fixed. Same model, same prompt, with and without the skill.

1 of 4

Prompt

Provide the complete terminal command to evaluate .agents/skills/frontmatter-repair against holdout fixtures and save evaluation results.

What was checked, without and with the skill
Checked	Without skill	With skill
Executes node .agents/skills/observable-receipt-distiller/scripts/evaluate-candidate.mjs with --candidate .agents/skills/frontmatter-repair, --holdout holdout, and --summary reports/evaluation-summary.json	Missed	Passed

case-19 · gemini-3.6-flash · 8/13/2026 · answers are capped and redacted

What's inside

The instructions your model reads when the skill loads.

SKILL.md25 lines8 filesincludes scripts

Distill a candidate skill only from observable receipts.

Read the distillation contract.
Use the training receipts under receipts/training/; reject marker files under receipts/negative/ according to the contract.
Generate a candidate skill in .agents/skills/frontmatter-repair/ by running:

node .agents/skills/observable-receipt-distiller/scripts/distill-candidate.mjs \
  --receipts receipts/training \
  --candidate .agents/skills/frontmatter-repair \
  --summary reports/distillation-summary.json

Evaluate the candidate on the 4 holdout fixtures:

node .agents/skills/observable-receipt-distiller/scripts/evaluate-candidate.mjs \
  --candidate .agents/skills/frontmatter-repair \
  --holdout holdout \
  --summary reports/evaluation-summary.json

Run npm test.
Do not modify any source receipt under receipts/.
Return a structured report only after the candidate exists, the holdout has run, and tests are complete.

Track record

Benchmark

7 / 24

test cases passed with the skill

Adoption

Users: 0
Times used: 0

Safety and source details

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026

View 15 scanner checks

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

learnprompt / learnprompt

research/articles/thinking-distillation-boundary/showcase/observable-receipt-distiller/fixture/.agents/skills/observable-receipt-distiller/SKILL.md

View on GitHub ↗

License NOASSERTIONImported 8/3/2026Version v1Bundled 7 files · scripts

More skills from the same source and nearby in the registry, with their headline benchmark lift.