The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and synthesizes untrusted content from external sources.
Ingestion points: The skill reads PR titles via gh search prs and X post content via the X.AI x_search tool.
Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands when the agent processes the retrieved text for the digest article.
Capability inventory: The agent has the ability to write files to output/articles/, execute shell commands like gh and jq, and send notifications via the local ./notify script.
Sanitization: No sanitization or filtering of the retrieved PR titles or tweet content is performed before synthesis.
[COMMAND_EXECUTION]: The skill makes extensive use of shell commands and local binaries for its core logic.
Evidence: It uses the gh CLI for GitHub interactions, jq for data processing, and local scripts ./secretcurl and ./notify to handle API requests and user alerts.
[EXTERNAL_DOWNLOADS]: The skill interacts with well-known external APIs to fetch activity data.
Evidence: It performs network operations targeting the GitHub API and the X.AI API (api.x.ai) using provided environment secrets (GH_GLOBAL, XAI_API_KEY).

shiplog