The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run a shell script ~/.claude/hooks/memory-search.sh (referenced in SKILL.md). This provides the agent with host-level execution capabilities to interact with its memory management system.
[DATA_EXFILTRATION]: The skill accesses sensitive local directories ~/.claude/sessions/, ~/.claude/memory/, and ~/.claude/skills/ (referenced in SKILL.md), which store interaction history, user prompts, and extracted knowledge.
[PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection by searching and recalling content from previous sessions which may contain untrusted or malicious instructions.
Ingestion points: History logs and memory files in ~/.claude/ (referenced in SKILL.md).
Boundary markers: No delimiters or instructions to ignore embedded commands were found.
Capability inventory: Execution of the memory-search.sh script (referenced in SKILL.md).
Sanitization: No input validation or output sanitization of retrieved session data is described.

claudeception-memory