The Agent Skills Directory

[PERSISTENCE]: The skill provides the /aprende-enable-hooks command which modifies the agent's global configuration file (~/.claude/settings.json) to install PostToolUse and Stop hooks. This configuration ensures that the skill's signal collection logic executes automatically after every tool use and at the end of each session.
[INDIRECT_PROMPT_INJECTION]: The skill scans conversation transcripts for specific patterns to generate new instructions and documentation (such as CLAUDE.md and AGENTS.md). This creates a vulnerability where malicious or deceptive content in a conversation could be promoted into the agent's permanent knowledge base and influence future sessions.
Ingestion points: Reads the full conversation history and internal signal files to extract learning candidates.
Boundary markers: None implemented; detection relies on keyword matching within the raw transcript.
Capability inventory: The skill has permissions to Write, Edit, and use Bash to modify local project files and global agent settings.
Sanitization: The skill relies on human-in-the-loop validation (Pass D) rather than programmatic sanitization to prevent the persistence of malicious data.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform environment management tasks, such as creating project memory directories (mkdir -p), identifying the project path (pwd), and cleaning up session signal files.

aprende