The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to install the fec-pp-cli tool using npx and go install from repositories and packages owned by the vendor (mvanhorn). These are documented as necessary components for the skill's functionality (SKILL.md).
[INDIRECT_PROMPT_INJECTION]: The skill implements a 'learning loop' where the agent's actions are guided by 'playbooks' and 'notes' retrieved from a local store using natural language queries. This architecture creates a vulnerability surface where data in the local store could influence agent behavior.
Ingestion points: Data enters the agent context via the JSON response from the fec-pp-cli recall command and is added to the store via teach and playbook amend operations (SKILL.md).
Boundary markers: The skill provides a 'Quoting rule' (SKILL.md) that instructs the agent to use single quotes and escape internal quotes to prevent command injection during shell expansion.
Capability inventory: The agent uses the Bash tool to execute commands derived from the local store.
Sanitization: The agent is explicitly instructed to strip PII (Personal Identifiable Information) from queries before they are committed to the learning store (SKILL.md).
[DATA_EXFILTRATION]: The CLI tool supports a --deliver webhook:<url> flag, allowing the agent to route the output of its commands to any external URL provided. This represents a potential exfiltration vector if the sink is directed to an unauthorized endpoint.
[DYNAMIC_EXECUTION]: Through the 'playbook' system, the agent is instructed to dynamically execute sequences of commands with slot substitution based on logic retrieved from the CLI's internal store at runtime (SKILL.md).
[COMMAND_EXECUTION]: The skill relies on the Bash tool to perform all operations, including tool installation, authentication setup, and executing the various FEC data retrieval and learning loop commands.

pp-fec