The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads internal application metadata from ~/.claude/memento/metrics.json and .claude/memento/metrics.json. These files contain performance statistics and skill metadata relevant to the agent's operation. This access is local and intended for the skill's primary reporting purpose.
[INDIRECT_PROMPT_INJECTION]: The skill processes skill names and metrics which could potentially be influenced by external sources. While this constitutes a potential attack surface, the risk is mitigated as the skill only generates a report and suggests commands to the user.
Ingestion points: ~/.claude/memento/metrics.json and .claude/memento/metrics.json (SKILL.md)
Boundary markers: Absent
Capability inventory: None (no subprocess calls, network ops, or file-writes detected)
Sanitization: Absent

memento-stats