The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from code diffs, pull requests, and source files via commands like git diff and git show. It lacks explicit instructions or boundary markers to ignore potential malicious prompts embedded within code comments or strings in the reviewed files.
Ingestion points: File SKILL.md defines triggers for reviewing diffs, PRs, and specific files using git commands.
Boundary markers: Absent. There are no specific delimiters or instructions to prevent the agent from obeying natural language instructions found within the code being analyzed.
Capability inventory: The skill is authorized to modify files (Step 6) and execute shell commands including git, bun typecheck, and cargo check.
Sanitization: None. The agent directly parses the content of the code changes to produce a findings report.
[COMMAND_EXECUTION]: The instructions direct the agent to perform shell operations to retrieve code context and verify fixes.
Evidence: The agent is instructed to use git diff, git show, bun typecheck, and cargo check to perform its tasks. While these are standard development tools, they represent an execution surface that could be leveraged if an indirect prompt injection succeeds.

maple-otel-spec-review