cnb-token
Warn
Audited by Gen Agent Trust Hub on Aug 21, 2026
Risk Level: MEDIUMCREDENTIALS_UNSAFECOMMAND_EXECUTION
Full Analysis
- [DATA_EXPOSURE]: The skill handles sensitive Personal Access Tokens (PAT) and directs the user to paste them into the chat environment. Evidence: Instructions in SKILL.md asking users to 'copy and paste the token back to you'.
- [DYNAMIC_EXECUTION]: The skill generates shell commands by concatenating user-supplied input into command strings. Evidence: SKILL.md contains templates like 'setx CNB_TOKEN ""' where is replaced by user input.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input (the token) and interpolates it into executed commands without sanitization or safety boundaries. Evidence: Lack of validation logic for the placeholder in SKILL.md which could allow shell metacharacters to alter command behavior.
- Ingestion points: User-provided token string in SKILL.md.
- Boundary markers: Absent; the token is directly placed inside double quotes in shell templates.
- Capability inventory: Shell execution (setx, export, git config, cnb command line tool).
- Sanitization: Absent; no logic is provided to escape shell metacharacters in the token.
Audit Metadata