The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted job listings and details from the SEEK API and incorporates user queries into a persistent local knowledge base via a 'learning loop' (commands such as teach, recall, and playbook amend). This creates an attack surface where external content could influence future agent behavior.
Ingestion points: Reads live job listings and descriptive content from au.seek.com.
Boundary markers: The instructions do not define explicit delimiters or 'ignore' instructions for the agent when processing external results.
Capability inventory: The skill allows for local file writing (--deliver file), network transmission (--deliver webhook), and modification of the local knowledge base.
Sanitization: Relies on the agent's manual compliance with instructions to strip personal identifiers before teaching queries.
[DATA_EXFILTRATION]: The skill provides an output delivery mechanism via the --deliver webhook:<url> argument, which permits the agent to POST command results—including potentially sensitive job history or account data—to an external endpoint.
[COMMAND_EXECUTION]: The skill's primary functionality is delivered through a custom binary (seek-pp-cli) which the agent is instructed to execute with various arguments.
[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install CLI and MCP components from the author's GitHub repository and via the npx package runner.

pp-seek