| name | source-command-autonomy-report |
| description | Show or verify a stored Session Autonomy Receipt (token/autonomy/cost) for a session. |
source-command-autonomy-report
Use this skill when the user asks to run the migrated source command autonomy-report.
Command Template
/autonomy-report — Session Autonomy Receipt viewer
Read-only viewer/verifier for the canonical Session Autonomy Receipts produced at
session finalization (ADR-0108). It does NOT regenerate receipts (finalization owns
generation); it renders or verifies what was stored, preserving the historical
pricing snapshot. Extends the CLI convention of token-report — it does not
duplicate it.
Usage
node contextkit/tools/scripts/autonomy-report.mjs --session <id>
node contextkit/tools/scripts/autonomy-report.mjs --session <id> --json
node contextkit/tools/scripts/autonomy-report.mjs --session <id> --verify
node contextkit/tools/scripts/autonomy-report.mjs --latest
node contextkit/tools/scripts/autonomy-report.mjs --all
node contextkit/tools/scripts/autonomy-report.mjs --all --mode subscription
What it shows
- Consumption mode (subscription / api / hybrid / unknown) and claim type
(measured / estimated / insufficient-evidence).
- Observed tokens, estimated baseline, token reduction, Autonomy Multiplier
(always labelled estimated/measured; a low-confidence result shows a range, never a
false-precision point), and additional autonomous capacity.
- API cost breakdown + cost-per-accepted-task (api/hybrid only). Subscription mode
never shows an invented dollar figure.
- Confidence, evidence basis, calibration id, and integrity status (
--verify
recomputes the hash and checks the Ed25519 signature when present).
Honesty
An estimate is always labelled estimated and carries its calibration basis
("pilot calibration — not a proven claim"). No multiplier is shown without a matched
calibration baseline or a real direct A/B comparison.