The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The report.mjs script accesses sensitive authentication data from ~/.openkbs/projectJWT and the OPENKBS_API_KEY environment variable. This is used to authenticate requests to the vendor's Project API.
[DATA_EXFILTRATION]: The skill transmits the resolved credentials and project metadata to https://project.openkbs.com. This is a vendor-owned domain used for its intended purpose of retrieving AI usage statistics.
[COMMAND_EXECUTION]: The skill uses execSync to run grep commands for scanning the project directory for compliance markers like disclosure components and human review workflows. This occurs in report.mjs using static search patterns.
[PROMPT_INJECTION]: The skill instructions (Step 2) require the agent to ingest and analyze untrusted project data from functions/ and site/ directories. This presents an indirect prompt injection surface where malicious content in project files could attempt to influence the compliance report's findings or the agent's behavior.
Ingestion points: functions/ directory, site/ directory, and the ai-act-report-data.json intermediate file.
Boundary markers: None are specified in the instructions for the analysis phase.
Capability inventory: The skill can execute shell commands via node report.mjs and write files to the project root.
Sanitization: No explicit sanitization or filtering of the ingested project code is performed before analysis.

ai-act-report