The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to modify the file system by applying fixes to other skills' code and frontmatter. It also interacts with Git to create branches and diffs for repairs within repositories.
[EXTERNAL_DOWNLOADS]: The skill uses automated web fetching to verify the status and schema of URLs referenced in the target skills' documentation and configuration.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it ingests untrusted data from external sources.
Ingestion points: Reads SKILL.md files and error_output logs from other (potentially malicious or compromised) skills.
Boundary markers: None identified in the provided instructions to distinguish between the repair instructions and data content.
Capability inventory: Includes file-write access (for patching skills), network-read access (via WebFetch), and Git command execution.
Sanitization: No specific sanitization or validation of the input content is documented before processing or applying code changes.
[SAFE]: The skill installation and setup processes reference official repositories associated with the author (BankrBot).

aeon-skill-repair