The Agent Skills Directory

[COMMAND_EXECUTION]: The automation script scripts/main.py uses subprocess.run to execute git commands for repository lifecycle management, including cloning, committing, and pushing code. The arguments are passed as lists to prevent shell injection, though the branch names and commit messages are derived from external GitLab data.
[COMMAND_EXECUTION]: The setup workflow in SKILL.md instructs the setup agent to perform several shell operations, including creating directories, compiling Python code for syntax verification, and using tar and curl to package and upload the automation to the backend service.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it instructs the agent to read and process content from GitLab issues and discussions, which are untrusted external data sources.
Ingestion points: Data enters the context via the _get_issue and _list_labeled_issues functions in scripts/main.py, as well as through direct curl calls performed by the agent to the GitLab API.
Boundary markers: The skill uses explicit boundary instructions in the agent's implementation prompt, warning that issue content is untrusted and instructing the agent to ignore any embedded malicious commands.
Capability inventory: The agent possesses the terminal and file_editor tools, has network access via curl and git, and is provided with the GITLAB_TOKEN secret.
Sanitization: The skill does not sanitize the raw issue data before the agent processes it, relying instead on the agent's instruction-level guardrails.
[CREDENTIALS_UNSAFE]: The skill manages the GITLAB_TOKEN secret, which is required for API access and repository operations. While the script attempts to mask the token in logs and injects it securely into the git environment, the fact that the token is forwarded to an agent that also processes untrusted input creates a potential exfiltration vector.
[EXTERNAL_DOWNLOADS]: The skill performs external network operations to fetch data from the GitLab API and clones source code from GitLab repositories to the local environment for processing.

gitlab-issue-to-mr