The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute curl commands against a local service at 127.0.0.1 to perform health checks and retrieve observation data.
[DYNAMIC_EXECUTION]: Utilizes a node -e script snippet to dynamically resolve the worker port by reading local configuration from ~/.claude-mem/settings.json.
[INDIRECT_PROMPT_INJECTION]: The skill processes observation data which represents an attack surface for indirect prompt injection.
Ingestion points: Fetches data from the worker API using search, timeline, and get_observations endpoints described in SKILL.md.
Boundary markers: Uses the formatCcsAlignLine utility which applies «…» fencing and a 500-character truncation limit to data before caching.
Capability inventory: Contains capabilities for atomic file writes to middle.jsonl, management of cursor.json, and conditional patching of local agent profile files.
Sanitization: Implements a shared formatter to strip tag framing and invisible characters from recalled text, and includes an 'exclude marks' system to purge specific records from the cache.

ccs-align