The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and analyze past AI session transcripts and project manifests. This creates a potential surface where malicious instructions embedded in historical data could attempt to influence the agent's logic or behavior during the review process.
Ingestion Points: Historical session turns are queried and processed as described in references/session-sources.md.
Capability Inventory: The skill can execute local CLI commands and file GitHub issues using the gh tool.
Boundary Markers: The skill relies on natural language instructions for the agent to 'paraphrase safe evidence' rather than using strict structural delimiters for ingested data.
Sanitization: The skill includes explicit instructions for a 'PII scrub' and a 'final privacy pass' to redact sensitive identifiers before any data is output or shared.
Command Execution: The skill utilizes local CLI tools to collect diagnostic information and perform reporting tasks.
Evidence: The instructions in references/environment-fingerprint.md and references/reporting.md reference the use of maui --version and the gh CLI tool for filing issues.
Context: These tools are standard components for the developer environment targeted by the skill and are used to gather metadata and transmit requested feedback to the vendor's designated repository.

maui-devflow-session-review