The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions utilize bunx to download and execute the algal-skills package from the NPM registry.
Evidence: Verified in the shell commands provided in SKILL.md: bunx algal-skills run writing-audit ... and bunx algal-skills run writing-evaluate ....
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text from local files or direct input, creating a surface for potential indirect prompt injection attacks where malicious content in the audited files could influence the agent's output.
Ingestion points: The src path parameter (targeting files like README.md) and the <text|path> argument specified in the skill metadata.
Boundary markers: There are no specific delimiters or instructions defined to prevent the agent from following commands embedded within the text being audited.
Capability inventory: The skill is granted exec tool permissions to run the underlying auditing and evaluation processes.
Sanitization: The input text is processed without explicit sanitization or filtering to remove potential instruction-like content.

algal-writing