The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted data by grepping the user's codebase for authentication-related strings and identifiers. This creates a surface where malicious instructions in the code could influence the agent's migration plan.
Ingestion points: Multiple grep commands in SKILL.md that scan source files, environment files, and dependency manifests.
Boundary markers: No specific delimiters or instructions are provided to the agent to treat the scanned code as data rather than instructions.
Capability inventory: The agent can read files, write to the filesystem, and execute shell commands (npm, go, tsc).
Sanitization: Scanned content is not sanitized before analysis.
[COMMAND_EXECUTION]: The skill uses shell commands like grep, npm install, and go build to analyze the project and verify the migration. These are standard development operations.
[CREDENTIALS_UNSAFE]: The skill search patterns target sensitive environment variables such as STYTCH_SECRET and DESCOPE_MANAGEMENT_KEY, as well as configuration files like .env. This is intended to identify credentials for replacement during the migration process.
[EXTERNAL_DOWNLOADS]: The skill references official SDKs and documentation from Descope and Stytch hosted on GitHub and NPM. These are well-known and trusted sources for this migration use case.

stytch-to-descope