The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes market signals and pain points fetched from untrusted external sources like Reddit, Hacker News, and GitHub Issues. If an attacker controls content on these platforms, they could attempt to influence the agent's analysis or output.
Ingestion points: Data is retrieved via scripts/fetch.py and passed to the agent for clustering and synthesis in the SKILL.md workflow.
Boundary markers: The skill lacks explicit delimiters or instructions for the agent to ignore potentially malicious commands embedded within the fetched market data.
Capability inventory: The agent has the ability to execute Python scripts, write to the file system (e.g., /tmp and docs/), and perform network operations.
Sanitization: While the script truncates body excerpts, it does not perform sanitization to detect or neutralize prompt injection attempts in the external content.
[DYNAMIC_EXECUTION]: The SKILL.md file utilizes inline Python scripts executed via python3 << 'PYEOF' to manage data flow, parse JSON, and format the final report. This pattern involves the runtime generation and execution of code blocks to facilitate the skill's logic.
[EXTERNAL_DOWNLOADS]: The skill performs automated searches across various public web services. Notably, the scripts/fetch.py script uses ssl._create_unverified_context() to disable SSL certificate verification for its network requests, which increases the risk of man-in-the-middle attacks where data could be intercepted or manipulated.
[COMMAND_EXECUTION]: The skill frequently invokes the python3 interpreter to run its core data collection script (scripts/fetch.py) and perform auxiliary tasks like input validation and file system cleanup.

map-your-market