The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and analyze the content of SKILL.md files from other installed skills to detect trigger overlaps and instruction conflicts. This creates a data ingestion surface where a malicious or poorly-configured skill already present in the environment could provide instructions that influence the auditor's conclusions or the agent's logic.
Ingestion points: The instructions in SKILL.md and SKILL.zh.md require the agent to read every discovered SKILL.md file in the environment.
Boundary markers: There are no explicit instructions to use delimiters or ignore embedded commands when processing the external skill content.
Capability inventory: The skill executes a local Node.js script (scripts/birdview.mjs) and writes assessment files to the .birdview/ directory.
Sanitization: No sanitization or validation of the processed instruction text is described.
[COMMAND_EXECUTION]: The skill relies on executing a local script (node <skill-root>/scripts/birdview.mjs) to perform the inventory and generate reports. This is a functional requirement for the skill's stated purpose but involves the execution of code included in the skill package.

birdview-compatibility