The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface because it is designed to scan project source code and extract information to generate documentation.
Ingestion points: The re-sync command (described in references/re-sync.md) performs a full project source scan, and the sync command (described in references/git-sync.md) reads content from git diff output.
Boundary markers: There are no explicit instructions or delimiters mentioned to prevent the agent from obeying instructions embedded within source code comments or strings during the scanning process.
Capability inventory: The skill has the capability to write multiple files to the filesystem (eo-doc/ directory) and, most critically, to modify CLAUDE.md (as detailed in references/claude-injection.md), which is used to provide the agent with global project context.
Sanitization: The analysis did not find any specific sanitization or filtering logic to strip potential malicious instructions from the source code content before it is interpolated into the generated documentation or the CLAUDE.md file.
[COMMAND_EXECUTION]: The skill provides the agent with specific shell commands to execute for its core functionality.
Evidence: references/git-sync.md instructs the use of git diff, git log, and jq to manage synchronization state and identify changed files. These are standard operations for a development-focused skill and do not involve administrative privileges.

eo-doc-manager