The Agent Skills Directory

[PROMPT_INJECTION]: No malicious prompt injection or safety bypass patterns were detected. The instructions use authoritative language to enforce accuracy and verification procedures, which serves as a safety-enhancing mechanism rather than a bypass.
[DATA_EXFILTRATION]: No data exfiltration patterns were found. The skill does not perform network operations or access sensitive system files. Its scope is limited to project files via the 'Read' tool.
[CREDENTIALS_UNSAFE]: No hardcoded secrets, API keys, or private credentials were found in the metadata or instructions.
[REMOTE_CODE_EXECUTION]: There is no remote code execution. The skill contains no shell commands, external script downloads, or dynamic execution patterns.
[EXTERNAL_DOWNLOADS]: No external URLs or remote resources are referenced for download or execution.
[OBFUSCATION]: The content consists of clear, plain-text markdown instructions. No hidden characters, Base64 encoding, or homoglyph substitutions were detected.
[INDIRECT_PROMPT_INJECTION]: While the skill instructs the agent to process project data, it does so to verify existence and state (e.g., checking if a file exists before claiming success). It does not exhibit vulnerabilities where untrusted data could override the agent's core instructions.
[COMMAND_EXECUTION]: The skill does not execute system commands. The 'allowed-tools' section in the frontmatter restricts the agent to the 'Read' tool only.

verification-before-completion