The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to handle external data that could contain malicious instructions, such as API responses, deep-link parameters, and push notification payloads.
Ingestion points: Untrusted data enters the agent's context through API integration, deep-link handling, and push notifications as specified in SKILL.md.
Boundary markers: The skill includes explicit instructions to "Treat API/deep-link/push payloads as untrusted" and to "validate all external payloads."
Capability inventory: The agent has the authority to write code, implement network logic, and execute build and test commands.
Sanitization: Instructions prioritize validation and state that client-side checks should not be treated as security boundaries.
[SAFE]: The skill demonstrates several security-positive patterns.
It mandates the use of platform-specific secure storage (Keychain/Keystore) for credentials.
It explicitly forbids the logging of secrets, tokens, or personal data and requires redaction in crash reports.
It includes a strict rule in references/runtime-capability-fallbacks.md against the automatic installation of skills, SDKs, packages, or native modules, which mitigates supply chain and remote code execution risks.

squad-mobile