The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize untrusted data from external sources such as pull requests, issues, feedback, and error logs. This data is not under the skill's control and could contain malicious prompts (e.g., in issue comments or PR descriptions) aimed at manipulating the agent's behavior during the summarization process.\n\t- Ingestion points: The skill reads pull requests, issues, feedback, errors, and telemetry from configured repositories and sources defined in .agent-factory/config.yaml.\n\t- Capability inventory: The skill is restricted to read-only reporting. It explicitly forbids taking actions such as replying, approving, merging, assigning, or closing based on the processed data.\n\t- Boundary markers: The skill lacks explicit delimiters or instructions for the agent to ignore embedded commands within the ingested text, relying on general instructions to follow configuration boundaries.\n\t- Sanitization: There are no instructions for sanitizing, escaping, or validating the content of the external records before they are processed and summarized.

factory-human-digest