The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon instructions from an external plan file provided as a command-line argument.
Ingestion points: The file path passed via $ARGUMENTS is read and processed by the agent to guide the implementation (/build-phase) and review phases.
Boundary markers: While the skill emphasizes critical thinking and re-reading the plan from disk, it does not specify the use of data delimiters or explicit 'ignore embedded instructions' prompts to separate the workflow's control logic from potentially adversarial content in the plan file.
Capability inventory: The skill has access to powerful tools including Bash, Write, Edit, and Agent, which could be exploited if a malicious plan file contains embedded commands or instructions to manipulate the filesystem.
Sanitization: There is no evidence of content sanitization or validation to ensure the plan file does not contain malicious command injections intended for the Bash tool.

build-model