teleport-access-review
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from
tctl access-review(identity names, resource names, aliases) which could contain malicious instructions. The skill proactively mitigates this inreferences/SECURITY.mdby instructing the agent to treat all output as untrusted data, avoid following instructions in the output, and sanitize/quote values before reuse in shell arguments. - Ingestion points:
tctl access-reviewcommand output (JSON/Text). - Boundary markers: Explicitly defined in
references/SECURITY.md(treat output as data only). - Capability inventory: Subprocess calls for
tctlandtsh; read-only post-processing withjq. - Sanitization:
references/SECURITY.mdmandates quoting and literal treatment of discovered data. - [SAFE]: The skill implements several best practices for security-sensitive administrative tools, including verifying active cluster status (
tsh status) and requiring human confirmation for any write/remediation actions.
Audit Metadata