The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from untrusted external sources, which could potentially contain malicious instructions intended to influence the agent's behavior during analysis.
Ingestion points: scripts/citability-scorer.py fetches HTML content from arbitrary URLs provided as arguments; scripts/llms-txt-gen.py retrieves file contents from GitHub repositories via the GitHub API; scripts/weekly-cite-check.py processes text responses returned from various AI engine APIs (Claude, GPT, Perplexity, Gemini, DeepSeek).
Boundary markers: There are no explicit boundary markers or instructions to ignore embedded prompts within the retrieved content before it is parsed or summarized.
Capability inventory: The skill uses urllib.request to perform network operations. It reads local files and outputs data to stdout, but does not utilize dangerous execution functions like eval() or exec() on the retrieved data.
Sanitization: While the skill uses HTMLParser to navigate document structure, it does not sanitize the text content of the documents or AI responses for potential prompt injection patterns before inclusion in reports.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to various external services to fulfill its auditing and optimization functions.
AI Service APIs: scripts/weekly-cite-check.py communicates with api.anthropic.com, api.openai.com, api.perplexity.ai, generativelanguage.googleapis.com, and api.deepseek.com to check for domain citations.
GitHub API: scripts/llms-txt-gen.py interacts with api.github.com to fetch repository content for generating llms.txt files, identifying the author's own repository (Gingiris-1031/growth-tools) as the default target.
Arbitrary URLs: scripts/citability-scorer.py fetches HTML from user-provided URLs to score them for citability as part of its core SEO audit functionality.

gr-geo-cite