* Brute-force search (fallback)

Exact top-k by cosine (brute force) — returns array of ids.

* Build dependency graph

* Build distance matrix from vertices

* Build the search query SQL.

Compact cache key — hash + length so distinct prompts of similar shape collide rarely.

* Write to TieredCache after DB write.

* Calculate capability match score

* Calculate secure checksum for session data using SHA-256

* Calculate secure checksum for session data using SHA-256

* Calculate confidence based on success rate

* Calculate Conditional VaR

* Calculate embedding similarity score

* Calculate priority for a file based on path and coverage

* Calculate schedule (earliest start, latest finish, float)

* Calculate text similarity (simplified)

* Calculate utilization based on claims and their priorities

* Calculate Value at Risk

* Calibrates the quantizer using sample vectors. * * @param samples - Representative vectors for calibration

* Check if agent can execute a task type

* Check if a plugin can be safely removed (no dependents).

* Canonical bytes for signing = `JSON.stringify(body)` UTF-8 encoded. * * Matches the plugin-registry signer (publish-registry.ts:127-151) and the

Canonical JSON: recursive key-sorting for deterministic serialization.

* Capture current performance as baseline

* Capture current contracts as baseline

* Check bucket and refill tokens

* Check agent health

* Check for drug interactions

* Check if state should change based on timeout

* Chunked attention - process in blocks

* Clamp a number to the range [min, max].

The classifier the SKILL.md documents — re-implemented here as the test contract.

* Classify market regime from price data