The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill contains references to a blacklisted and potentially malicious URL within its source discovery metadata.
Evidence: Found in _workspace/discovery/frameworks.json and _workspace/discovery/interviews.json.
URL: https://altair.pw/pub/lib/Martin%20Heidegger%20-%20Being%20and%20Time%20(translated%20by%20Macquarrie%20&%20Robinson).pdf.
Risk: The domain altair.pw has been flagged as malicious by automated scanners. The inclusion of links to untrusted and blacklisted domains in the skill's internal knowledge base poses a risk if the agent or user attempts to access these resources.
[PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection due to its primary purpose of interpretive analysis on untrusted input.
Ingestion points: The skill is explicitly triggered by and processes user-provided discussions about technology, mortality, and data-driven worldviews (SKILL.md).
Boundary markers: Absent. There are no instructions to use delimiters or ignore instructions embedded in the user-provided text.
Capability inventory: Low risk. While the agent interprets input, the skill itself does not provide the agent with dangerous capabilities like subprocess execution, file system writes, or active network operations.
Sanitization: Absent. The skill instructions do not require the agent to filter or sanitize external content before applying Heideggerian frameworks.

martin-heidegger