Look up the trust verdict for a counterparty DID. Never raises on a network/parse failure — returns an `unknown` verdict instead, leaving the

* Build context progress bar with ANSI colors. * @param {number} remaining - Raw remaining percentage from Claude Code * @returns {string} Colored b

* Check if a command requires governance approval.

* Detect recurring failure patterns from skill runs. * * @param {Array} skillRuns - Array of skill run objects (newest first) * @param {Object} [op

* Extract modified file paths from tool input.

* Walk up from `startDir` until a directory containing a known project * root marker (package.json or formatter config) is found. * Returns `startDi

* Get the Claude config directory (alias of getAgentDataHome for backwards compatibility).

* Load aliases from file * @returns {object} Aliases object

Load all instincts: project-scoped + global. Project-scoped instincts take precedence over global ones when IDs conflict.

Parse a JSONL observation trace file into sorted events.

* Read cumulative cost for a session from costs.jsonl. * * Scans the full file because each row is a cumulative session total * (see cost-tracker.j

* Build a compressed catalog from a directory of agents. * * Modes: * - 'catalog': name, description, tools, model only (~2-3k tokens for 27 agent

* Create a JSONL transcript file with N user messages. * Each line is a JSON object with `"type":"user"`.

* Format duration from ISO timestamp to now. * @param {string} isoTimestamp * @returns {string} e.g. "5s", "12m", "1h23m"

* Get the bridge file path for a session. * @param {string} sessionId - Already-sanitized session ID * @returns {string}

* Get the package manager to use for current project * * Detection priority: * 1. Environment variable CLAUDE_PACKAGE_MANAGER * 2. Project-specifi

Parse a YAML compliance spec file.