The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a significant ingestion surface for untrusted data, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The skill performs tool calls to Reddit (reddit.com/search.json), Hacker News (Algolia API), and general web search/fetch tools as defined in SKILL.md to retrieve posts, comments, and article content.
Boundary markers: While the skill enforces strict citation rules and audit logging, it does not explicitly instruct the agent to use secure delimiters or specific 'ignore instructions' markers when interpolating the fetched untrusted content into the prompt.
Capability inventory: The skill allows the agent to execute local Python scripts, write research briefings to ${RESEARCH_DIR}/pulse/, and maintain session metadata in ~/.pulse_sessions/.
Sanitization: The provided utility scripts (citation_tracker.py, time_window_calculator.py, etc.) handle JSON structure and arithmetic but do not sanitize or filter the natural language content of the retrieved data for prompt injection attacks before processing.
[COMMAND_EXECUTION]: The skill requires the agent to execute three local Python scripts included in the package (scripts/citation_tracker.py, scripts/time_window_calculator.py, and scripts/topic_slug_generator.py).
These scripts are used for essential functions such as computing search timestamps, generating filesystem-safe slugs, and managing a session audit log.
The scripts interact with the local filesystem, creating directories and files in the user's home directory (~/.pulse_sessions and ~/research). All scripts utilize only the Python Standard Library.

pulse