The Agent Skills Directory

[PROMPT_INJECTION]: The skill documentation contains an explicit instruction to bypass security scanning mechanisms.
Evidence: In the 'Pitfalls' section of SKILL.md, it suggests a "Fix" for scanner false positives by marking jobs with skip_injection_scan: true. This instruction attempts to influence the platform's security posture by disabling safety filters.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and processing untrusted data from multiple external sources.
Ingestion points: The skill scrapes content from GitHub trending, Hacker News, and arXiv (SKILL.md, Workflow Step 1).
Boundary markers: There are no instructions or delimiters provided to ensure the AI ignores malicious instructions that might be embedded in the scraped tech news or paper titles.
Capability inventory: The skill has the capability to write files to the local workspace (SKILL.md, Workflow Step 3), perform network requests to external TTS APIs (Step 5), and delete files using Python's Path.unlink() (Step 6/Rules).
Sanitization: No sanitization, escaping, or validation logic is described for the content retrieved from external sources before it is passed to the AI summarization step.

daily-briefing