The Agent Skills Directory

[COMMAND_EXECUTION]: The primary function of this skill is to provide instructions and a full reference for executing subcommands of the erk exec CLI tool. These commands facilitate workflows such as PR review management, plan operations, and inter-process communication.
[EXTERNAL_DOWNLOADS]: Several commands involve fetching data from external remote sources:
download-remote-session: Downloads session data from GitHub gists or specific git branches.
fetch-sessions: Downloads preprocessed sessions from remote planned-pr-context branches.
[PROMPT_INJECTION]: The skill exhibits a vulnerability surface for indirect prompt injection because it ingests untrusted data from external sources and provides it to the agent's context.
Ingestion points: Commands such as get-pr-review-comments, get-issue-body, get-pr-discussion-comments, get-pr-feedback, and get-review-activity-log (found in reference.md) fetch content directly from GitHub PRs and issues.
Boundary markers: No specific delimiters or warnings to ignore embedded instructions were identified in the command documentation.
Capability inventory: The skill possesses significant capabilities across its command set, including file system operations (marker create/delete), network uploads (push-session), and the ability to modify GitHub state (update-issue-body, post-pr-inline-comment, close-pr).
Sanitization: There is no evidence of sanitization, filtering, or validation of the content fetched from GitHub before it is used to inform agent actions.

erk-exec