The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and processing untrusted data from external URLs and local project files to generate SEO reports and AI-readiness scores.\n
Ingestion points: Commands such as /seo audit and /seo page ingest content from arbitrary URLs and local paths provided by the user, which are then parsed by scripts like parse-meta.ts and analyze-keywords.ts mentioned in SKILL.md.\n
Boundary markers: There are no explicit instructions or delimiters defined in the orchestrator to prevent the agent from following instructions embedded within the processed web content.\n
Capability inventory: The skill possesses the ability to write analysis reports to the local filesystem (.fuse-seo/reports/), create marker files at the project root (touch .fuse-seo), and delegate tasks to sub-agents.\n
Sanitization: The skill lacks documented evidence of sanitization or filtering of the ingested content before it is submitted to LLM-based scoring tools like geo-score.ts.\n- [COMMAND_EXECUTION]: The orchestrator utilizes local system tools to perform technical audits.\n
Evidence: SKILL.md references scripts that wrap command-line utilities, including check-cwv.ts (a Lighthouse CLI wrapper) and diff-seo.ts (which executes git diff for drift analysis).

seo