Name: ledger
Author: belemcrizan

Imported from belemcrizan/Aeris_Competition (submission/skills/ledger/SKILL.md). Install upstream with npx skills add belemcrizan/Aeris_Competition --skill ledger. Copyright stays with the author.

ledger

State lives in /tmp/aeris/ledger.json (never in /workspace). One task per sandbox, so no reset is needed; ledger.py reset clears it.

Commands (script: scripts/ledger.py)

claim "text": record a fact extracted from the issue.
add H1 --target sym --weight 0.5 --claim "mechanism": add a hypothesis.
update H1 [--weight W] [--for "evidence"] [--against "evidence"] [--reject]: update one.
experiment --command "cmd" --result "outcome" [--purpose "why"]: record an investigation step.
patch --files a.py,b.py --hypothesis H1 --result pass|fail|untested [--note "why"]: record an attempt.
status [--budget-used 0.4]: print the ledger, normalized weights, entropy, uncertainty level, budget mode and recommended action.

Output

Every command prints the compact ledger. status adds lines such as:

MODE: NORMAL  UNCERTAINTY: MEDIUM  top=H1 p=0.55 entropy=0.62
ACTION: DISCRIMINATE H1 vs H2 with the cheapest check whose outcome differs

Weights are confidence weights, not calibrated probabilities.

{ "ocm": "1", "id": "belemcrizan-aeris-competition-ledger", "kind": "skill", "name": "ledger", "description": "Compact evidence ledger for one task. Tracks issue claims, competing root-cause hypotheses with confidence weights, experiments and patch attempts, and reports entropy-based uncertainty and a budget-aware recommended action.", "publisher": "belemcrizan", "version": "1.0.0", "capabilities": { "domains": [ "finance", "science" ], "tags": [ "skill-md", "github" ], "languages": [ "en" ] }, "quality_prior": 0.6, "examples": [ "Compact evidence ledger for one task. Tracks issue claims, competing root-cause hypotheses with confidence weights, experiments and patch attempts, and reports entropy-based uncertainty and a budget-aware recommended action." ], "primary": false, "metadata": { "source": { "provider": "github", "repository": "https://github.com/belemcrizan/Aeris_Competition", "path": "submission/skills/ledger/SKILL.md", "ref": "242db46024430a8fec504b09ad239e828406e84a", "url": "https://github.com/belemcrizan/Aeris_Competition/blob/242db46024430a8fec504b09ad239e828406e84a/submission/skills/ledger/SKILL.md", "key": "belemcrizan/Aeris_Competition/submission/skills/ledger/SKILL.md" } }, "instructions": "# ledger\n\nState lives in /tmp/aeris/ledger.json (never in /workspace). One task per sandbox, so no reset is needed; `ledger.py reset` clears it.\n\n## Commands (script: scripts/ledger.py)\n\n- `claim \"text\"`: record a fact extracted from the issue.\n- `add H1 --target sym --weight 0.5 --claim \"mechanism\"`: add a hypothesis.\n- `update H1 [--weight W] [--for \"evidence\"] [--against \"evidence\"] [--reject]`: update one.\n- `experiment --command \"cmd\" --result \"outcome\" [--purpose \"why\"]`: record an investigation step.\n- `patch --files a.py,b.py --hypothesis H1 --result pass|fail|untested [--note \"why\"]`:", "cost": { "context_tokens": 261 } }

ledger

About

ledger

Commands (script: scripts/ledger.py)

Output

Use it

Manifest

Reviews