The Agent Skills Directory

[PROMPT_INJECTION]: Vulnerability to indirect prompt injection from data ingested via the TrueFoundry API.
Ingestion points: API responses from TrueFoundry endpoints (e.g., secret lists, app statuses) and metadata from Git repositories (SKILL.md, references/api-endpoints.md).
Boundary markers: Absent; data from external sources is incorporated into the agent's context without explicit isolation markers.
Capability inventory: Shell command execution (Bash tool), network operations (curl via scripts/tfy-api.sh), and system configuration changes (tfy apply).
Sanitization: Absent; data retrieved from the API is used to construct deployment manifests and shell commands without explicit validation.
[COMMAND_EXECUTION]: The skill uses templates to dynamically generate YAML manifests based on user input and environment data, which are then processed by the tfy CLI or submitted to the REST API (references/manifest-defaults.md, references/cli-fallback.md).
[EXTERNAL_DOWNLOADS]: The skill contains logic to automatically install the truefoundry Python package and CLI tool from the standard Python Package Index (PyPI) if they are not already installed (references/prerequisites.md).

truefoundry-secrets