The Agent Skills Directory

[PROMPT_INJECTION]: The skill is subject to indirect prompt injection because it ingests and processes untrusted data from the local repository to make decisions about which agent capabilities to activate.
Ingestion points: Processes repository manifests (package.json, pyproject.toml, etc.) and documentation files (SKILL.md).
Boundary markers: Absent. No explicit delimiters or warnings are used to prevent the agent from obeying instructions embedded in the analyzed files.
Capability inventory: Reads files via shell commands and manages the installation/activation of other skills and rules.
Sanitization: None detected. The skill relies on direct evidence from file content without filtering.
[COMMAND_EXECUTION]: The skill uses local shell commands (rg, cat) to establish project context. While used legitimately for analysis, these tools provide the mechanism for the agent to access the file system based on potentially untrusted repository content.

agent-sort