The Agent Skills Directory

[PROMPT_INJECTION]: The skill has a vulnerability to indirect prompt injection due to its reliance on untrusted repository data. 1. Ingestion points: The skill reads service names, port numbers, and build configurations from docker-compose.yml, package.json, and other manifest files (Sections 2.1, 2.2, 2.3). 2. Boundary markers: No specific delimiters or safety instructions are used when interpolating discovered data into prompts or commands. 3. Capability inventory: The skill can execute shell commands like git commit, gh pr create, and okteto build, and it can write files to the repository root. 4. Sanitization: There is no evidence of validation or sanitization of the values extracted from the repository files before they are used in commands or the generated okteto.yaml.
[COMMAND_EXECUTION]: The skill executes shell commands using unvalidated data from the local repository, creating a risk of command injection. Evidence: Section 6.2 provides a shell script template for committing and pushing code where placeholders like are replaced with discovered service data. Evidence: Section 5.2 executes the okteto build command for services discovered in the file system.
[DATA_EXFILTRATION]: The skill is designed to push project configuration to a remote repository. Evidence: Section 6.2 uses git push and gh pr create to upload the new Okteto manifest to a remote branch and create a pull request. Context: This is the intended purpose of the skill and uses standard git/github tools.

okteto-onboarding