The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/status.sh executes several shell commands including git status, git log, git rev-list, grep, and sed to collect information about the workspace. While these are standard tools for status reporting, they involve executing sub-processes based on local file system content.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection by processing and displaying untrusted content from the local environment.
Ingestion points: Data is read from workspace configuration files (project names) and git repositories (commit messages via git log).
Boundary markers: None. The extracted strings are placed directly into a formatted table and list for the agent to process.
Capability inventory: The agent has the capability to execute shell commands and read/write files as part of its normal operation.
Sanitization: No sanitization or escaping is performed on the commit messages or project names before they are presented to the agent. An attacker who can influence a commit message in a watched repository could theoretically inject instructions into the agent's context.

workspace-status