AANA Purchase Booking Guardrail Skill
Analysis
This is an instruction-only purchase and booking safety guardrail that requires explicit approval and redaction before financial commitments.
Findings (2)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Checks for instructions or behavior that redirect the agent, misuse tools, execute unexpected code, cascade across systems, exploit user trust, or continue outside the intended task.
Use this skill when an OpenClaw-style agent may purchase, book, reserve, subscribe, renew, upgrade, downgrade, cancel, refund, bid, donate, transfer funds, or take any irreversible or financially binding action.
The skill applies to high-impact financial actions, but the surrounding instructions are guardrails requiring verification, reversibility checks, and explicit user approval before final submission.
Checks whether tool use, credentials, dependencies, identity, account access, or inter-agent boundaries are broader than the stated purpose.
Check payment privacy: do not expose full payment numbers, bank details, credentials, or unnecessary account data.
Purchase and booking workflows can involve payment methods, account details, and identity data; the artifact acknowledges this sensitive context and instructs redaction and minimization.
