Component Evidence

Component Evidence

Measure the package without pretending that installation or static structure proves usefulness. Combine two separate evidence sources:

1. A read-only Alex ACT Brain Compiler assessment provides structural importance for each skill.
2. An explicit local meditation inventory captures traceable skill use.
3. An explicit local outcome record captures whether a component helped a task.

Boundaries

• Inventory only skills explicitly traceable to the session and present in the configured static assessment. Do not infer usage from installation, filenames, incidental prose, or skills from another package.
• Record an outcome only when the user explicitly provides it.
• Store only a component identifier, event type or outcome, and timestamp. Do not store task text, prompts, source bodies, paths, identities, session IDs, or other private content.
• Static importance is a structural proxy, not semantic meaning, host discovery, invocation telemetry, or proof of usefulness.
• A hybrid score is unavailable until both static and outcome evidence exist.

Static Assessment

Run Brain Compiler's assessor against the package or another target. It does not change the target:

node <brain-compiler>\scripts\assess-brain.cjs --root <target-root> --out <outside-target>\assessment.json

For each skill, skillImportance reports a score out of 60 with its raw signals:

Treat these signals as an explainable integration measure. A low score can mean the skill is intentionally independent, not unimportant.

Configure The Shared Ledger

Configure the shared evidence root once on each device. The configuration is local to the device; point it to that device's synchronized view of the same OneDrive folder:

node <this-skill>/scripts/component-evidence.cjs configure --evidence-root <absolute-onedrive-path>
node <this-skill>/scripts/component-evidence.cjs configure --evidence-root <absolute-onedrive-path> --apply

Applied configuration is stored at ~/.scout/component-evidence.json. It contains only the selected evidence-root path. Later commands use this configured location when --evidence-root is omitted and fail rather than creating an inferred local ledger.

If a device has no configured evidence root, use scout-shared-data-setup to preview and approve the shared OneDrive scaffold before returning here.

Local Outcome Ledger

After configuration, bootstrap the ledger once:

node <this-skill>/scripts/component-evidence.cjs bootstrap
node <this-skill>/scripts/component-evidence.cjs bootstrap --apply

Record only an explicit user outcome:

node <this-skill>/scripts/component-evidence.cjs record --component <skill-name> --outcome <helped|neutral|not-helpful>
node <this-skill>/scripts/component-evidence.cjs record --component <skill-name> --outcome helped --apply

Every mutation previews by default. Applied records are newline-delimited JSON at <evidence-root>/component-evidence/events.ndjson.

Meditation Inventory

At meditation, inventory each unique skill explicitly traceable to the current session and present in the configured assessment.json. Do not list a skill merely because it was installed, mentioned, available, or supplied by another package. Preview the exact component names, then apply only after explicit approval:

node <this-skill>/scripts/component-evidence.cjs inventory --components <skill-name,skill-name>
node <this-skill>/scripts/component-evidence.cjs inventory --components <skill-name,skill-name> --apply

The command deduplicates names within one inventory pass. It records used events separately from outcomes; an inventory establishes usage evidence, not usefulness evidence.

Hybrid Report

Combine the static assessment and local outcome ledger:

node <this-skill>/scripts/component-evidence.cjs report

Each component reports:

• staticImportance: the Brain Compiler score and raw signals;
• usage: recorded meditation inventories and the latest use timestamp;
• usefulness: outcome counts, a smoothed 0-100 score, confidence based on record count, and the latest outcome timestamp; and
• hybridScore: 40% structural importance and 60% usefulness, only when both inputs exist.

The usefulness score applies a neutral prior so a single outcome cannot produce a misleading 0 or 100. Confidence is low for 1-4 outcomes, medium for 5-14, and high for 15 or more. Use the raw counts with the score.