The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several shell commands to facilitate research and notification.
Uses curl to query the Semantic Scholar API for academic paper metadata.
Uses gh api to fetch repository details, including commit history, pull requests, issues, and README files.
Uses a local ./secretcurl script to interact with the Replicate API for image generation, utilizing a placeholder for the API token to ensure secret safety.
Uses a local ./notify script to post article updates.
[EXTERNAL_DOWNLOADS]: The skill fetches data from several external services to fulfill its research purpose.
Queries api.semanticscholar.org for technical paper details.
Queries api.replicate.com for image generation services.
Interacts with github.com via the GitHub CLI to retrieve repository content.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, creating a potential attack surface for indirect prompt injection.
Ingestion points: Fetches content from external web pages via WebFetch, academic papers via Semantic Scholar, and repository data (READMEs, commit messages, PR titles) via GitHub.
Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the fetched content.
Capability inventory: The agent has the capability to execute shell commands (curl, gh api) and write files to the local output/ directory.
Sanitization: There is no evidence of specific sanitization or filtering applied to the external text before it is processed by the agent.

article