The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data from GitHub and X to synthesize its reports, creating a surface for indirect prompt injection.
Ingestion points: PR titles and metadata fetched via the GitHub API (Step 2) and social media posts/mentions fetched via the X.AI API or web scraping (Steps 3 and 4).
Boundary markers: The instructions lack explicit delimiters or specific warnings to the model to ignore instructions potentially embedded within the fetched PR descriptions or social media text during the synthesis step (Step 6).
Capability inventory: The skill has the ability to write files to the output/articles/ directory and send notifications using the ./notify script.
Sanitization: No explicit sanitization or filtering mechanisms for the external data are described before the content is processed for synthesis.
[EXTERNAL_DOWNLOADS]: The skill interacts with well-known technology services to fetch data.
Fetches activity data from the GitHub API using the gh command-line tool.
Fetches social media content and engagement metrics from the X.AI API (https://api.x.ai/v1/responses) using curl.
[COMMAND_EXECUTION]: The skill utilizes standard system and project-specific utilities to perform its operations.
Uses the gh CLI for repository searches and API interactions.
Uses curl (via a local secretcurl script) to communicate with the X.AI API.
Uses jq, date, and standard shell redirection for data processing and local state management in the memory/ directory.

shiplog