The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill creates a structural surface for indirect prompt injection by design, as it must ingest configuration values from environment variables and external secret stores to function.
Ingestion points: Values are retrieved from os.environ and cloud backends (GCP, AWS, Vault) in SKILL.md and references/secret-manager-integration.md to populate a centralized Settings object.
Boundary markers: Robust validation is implemented via Pydantic types (Literal, HttpUrl, SecretStr) and custom validators to ensure configuration values meet expected formats.
Capability inventory: The skill manages credentials and identifiers used for network requests to LLM providers, LangSmith, and Postgres databases as documented in references/startup-smoke-test.md.
Sanitization: Input is validated against schemas, but identifiers like prompt hashes are not further sanitized before use in service pull requests.
[EXTERNAL_DOWNLOADS]: The skill facilitates the retrieval of secrets and prompt configurations from well-known and trusted service providers.
Details: Fetches secrets from GCP Secret Manager, AWS Secrets Manager, and HashiCorp Vault using official SDKs in references/secret-manager-integration.md. It also pulls prompt templates from LangSmith using the official client library as described in SKILL.md.

langchain-multi-env-setup