The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes documentation links to 'jeffallan.github.io' which have been explicitly flagged as malicious by URL reputation scanners. Providing these links could lead users to phishing or malware-hosting sites.
[METADATA_POISONING]: The skill's metadata contains author URLs and documentation links associated with a blacklisted domain, which could be used to establish a false sense of trust or lead to malicious resources.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes user requirements to generate and execute code without sufficient sanitization or boundary markers.
Ingestion points: Component, state management, and routing requirements provided in the user's prompt.
Boundary markers: None present to distinguish untrusted data from instructions.
Capability inventory: Executes 'ng build' via subprocess (SKILL.md); performs file system writes to create components and service files.
Sanitization: None observed for user-supplied data used in code generation.
[COMMAND_EXECUTION]: The core workflow includes running shell commands like 'ng build --configuration production'. While a standard development tool, this capability could be exploited if malicious inputs influence the build process.

angular-architect