The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command to perform literature searches: python3 ../proposal-lit-search/scripts/search.py "the idea's core terms".
Evidence: Found in the literature grounding section of SKILL.md.
Risk: The arguments for this command are derived from user-provided "core terms." If these terms contain shell metacharacters (e.g., ;, |, &&), it could lead to arbitrary command injection on the host system.
[EXTERNAL_DOWNLOADS]: The skill fetches data from several well-known academic APIs to ground ideas in existing literature.
Evidence: Requests to api.crossref.org, dblp.org, and export.arxiv.org in SKILL.md.
Context: These are well-known scholarly services, and their use is consistent with the skill's primary purpose.
[REMOTE_CODE_EXECUTION]: The skill attempts to execute a script located in a sibling directory (../proposal-lit-search/scripts/search.py).
Risk: This execution pattern relies on code located outside the current skill's directory. While likely part of a modular skill set by the same author, this creates a dependency on unverified external files.
[PROMPT_INJECTION]: The skill processes external data from literature search results and user input, which are then used to influence agent behavior and file generation.
Ingestion: User input (thesis ideas) and API results (paper metadata).
Capability: Command execution (python3) and file writing (<slug>.md).
Sanitization: There are no explicit instructions for the agent to sanitize or escape results before processing them, though the Socratic instructions provide a conceptual boundary.

proposal-ideate