The Agent Skills Directory

[PROMPT_INJECTION]: The skill establishes a workflow that makes the agent susceptible to indirect prompt injection. By instructing the agent to read and debug external files using high-privilege tools, malicious instructions embedded within the processed code (e.g., in comments or string literals) could potentially influence the agent's behavior.
Ingestion points: Local file system access via the read and edit tools as specified in the Capabilities and Debugging sections of SKILL.md.
Boundary markers: Absent. The instructions do not include guidance for the agent to use delimiters or ignore embedded instructions when reading file content.
Capability inventory: Includes file system modification (write, edit, apply_patch) and arbitrary command execution (exec).
Sanitization: No sanitization or validation of the data ingested from the local file system is performed before processing.
[COMMAND_EXECUTION]: The skill explicitly directs the agent to use the exec tool to run commands such as npx tsc --noEmit and npm test. These operations are standard for a software engineering environment and utilize well-known development tools.
[NO_CODE]: The skill consists entirely of instructional markdown and YAML configuration without any accompanying scripts or executable binaries.

software-engineering