The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes Heroku CLI commands to inventory resources. This is strictly scoped to a pre-defined whitelist of read-only 'list' and 'info' commands as detailed in discover-live-capture.md. The skill includes explicit security logic to avoid capturing credentials or session tokens.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data from files such as Procfile, app.json, and Terraform configurations.
Ingestion points: discover-terraform.md, discover-live.md, and discover-billing.md ingest workspace files.
Boundary markers: SKILL.md contains an 'Input Security' section that explicitly instructs the agent to treat these files as data and ignore any embedded instructions or directives.
Capability inventory: The skill has file-writing capabilities, network access via designed AWS resources (for the user to deploy), and command execution through the Heroku CLI.
Sanitization: Configuration variables are filtered to capture keys only, redacting all values to prevent sensitive data leakage.
[DYNAMIC_EXECUTION]: The skill utilizes a structured _exec pattern defined in INTERPRETER.md to dispatch non-interactive work (like resource discovery) to isolated sub-agents. These sub-agents are assigned specific capability tiers (ro, rw, rwx) to enforce the principle of least privilege.
[DYNAMIC_CONTEXT_INJECTION]: SKILL.md uses the ! syntax (e.g., !uv --version) to perform environment checks at skill load time. This is used for benign tool detection to ensure necessary utilities like uv or python3 are available for the migration process.

heroku-to-aws