The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of prompt injection or safety bypass patterns was detected. The instructions are professional, focused on financial modeling, and lack any attempt to override agent constraints.
[DATA_EXPOSURE_AND_EXFILTRATION]: No sensitive data exposure or exfiltration patterns were identified. There are no hardcoded credentials, sensitive file paths, or network operations. The skill primarily provides templates for accounting calculations.
[OBFUSCATION]: No obfuscated content, such as Base64-encoded commands, zero-width characters, or homoglyphs, was found in any of the skill files.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns were detected. The skill does not perform external downloads or execute remote scripts.
[COMMAND_EXECUTION]: The skill does not contain any shell commands or subprocess execution patterns.
[PRIVILEGE_ESCALATION]: No privilege escalation attempts were found. The skill does not use administrative commands or attempt to modify system permissions.
[PERSISTENCE]: No persistence mechanisms, such as modifications to startup scripts or scheduled tasks, were identified.
[METADATA_POISONING]: The skill's metadata is consistent with its stated purpose and does not contain deceptive or malicious instructions.
[INDIRECT_PROMPT_INJECTION]: The skill defines a structured ingestion surface for financial data (pricing, SLA terms, revenue actuals). However, it does not include tools or scripts that execute operations on this data, reducing the risk of indirect injection attacks. The 'Capability and Permission Boundaries' section explicitly restricts the skill to read-only analysis and drafting.
[DYNAMIC_EXECUTION]: No dynamic code generation or execution patterns (e.g., eval, exec, or unsafe deserialization) were found.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not utilize dynamic context injection syntax or shell-replacement placeholders.

saas-agent-deferred-revenue-and-credit-reserves