The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill scans existing project files for documentation during the onboarding process, which presents a surface for indirect prompt injection if those files contain malicious instructions designed to influence the agent.
Ingestion points: The agent scans the existing repository tree for "doc-shaped content" (scattered notes, research, design docs) as described in EP3.
Boundary markers: The skill explicitly mandates a human-in-the-loop safety check, requiring the agent to "Produce a migration plan... show it to the user, move nothing until approved."
Capability inventory: The agent is instructed to perform file system operations, including creating directories, creating symlinks, and modifying configuration files like AGENTS.md or CLAUDE.md.
Sanitization: No specific sanitization or escaping of the scanned content is mentioned, as the skill delegates the final formatting and writing to other domain-specific context skills.
[COMMAND_EXECUTION]: The execution procedure involves standard file system modifications to establish the /ctx mount point, such as creating symlinks to external directories and appending pointer lines to project configuration files.

ctx-adopt