The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text from external web sources, which could contain malicious instructions designed to influence the agent's behavior during the expansion or clustering phases. \n
Ingestion points: In SKILL.md (Step 1), the skill ingests 'Related searches' and 'People Also Ask' (PAA) questions from WebSearch. In references/serp-overlap-methodology.md, it ingests URL sets and organic result data from search engines and the DataForSEO API. \n
Capability inventory: The skill possesses file-writing capabilities (creating .json plans, .md briefs, and .html visualizations) and can invoke the blog-write skill for automated content creation. \n
Boundary markers: The instructions do not define specific delimiters or instructional barriers to prevent the agent from following commands embedded in the retrieved search queries or PAA questions. \n
Sanitization: There is no explicit mechanism described to sanitize or filter the external text data before it is interpolated into prompts for subsequent tasks. \n- [COMMAND_EXECUTION]: The skill executes local Python scripts to facilitate its operations. \n
Evidence: SKILL.md and references/serp-overlap-methodology.md specify the execution of python scripts/dataforseo_costs.py and python scripts/fetch_page.py via subprocess calls. \n- [EXTERNAL_DOWNLOADS]: The skill retrieves data from external web services as part of its core functionality. \n
Evidence: It performs network requests to search engines (via WebSearch) and the DataForSEO API to obtain SERP data. It also references the related skill repository at github.com/AgriciDaniel/claude-blog for content execution.

seo-cluster