The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill generates shell scripts (scripts/validate-<name>.sh) at runtime based on documentation content and project code. It then executes these scripts to perform 'check-only' validation of runbooks. This involves the execution of dynamically generated code based on the project's state.
[COMMAND_EXECUTION]: The skill utilizes several shell commands, including git for repository synchronization (fetch, pull, stash, rev-parse) and bash to execute the generated validation scripts. This provides a mechanism for system-level interaction based on the interpreted code logic.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it establishes a 'ground truth' citation pool by reading the entire codebase, including source files, configurations, and existing documentation. This untrusted data is interpolated into the generated documentation and validation scripts.
Ingestion points: Reads all files in the repository to establish project facts (entry points, config, scripts, env vars, endpoints) and existing documentation (README.md, docs/*.md).
Boundary markers: The skill does not explicitly use boundary markers or delimiters when processing code content to prevent embedded instructions from influencing its behavior.
Capability inventory: The skill has capabilities to write files to the repository (docs/*.md, scripts/*.sh), execute shell commands (git, bash), and perform network connectivity checks (curl, nc within generated scripts).
Sanitization: There is no evidence of sanitization or filtering of the content read from the codebase before it is used to generate documentation or shell script logic.
[EXTERNAL_DOWNLOADS]: The skill references installation via npx skills and agent-skill-manager from the author's GitHub repository. It also performs git pull and git fetch operations to synchronize the repository with its remote origin before making changes. These resources are associated with the vendor's own infrastructure.

doc-manager