The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill requires the execution of a local shell script .claude/lib/checkpoint.sh via bash to record phase progress checkpoints.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill employs curl during the acceptance phase (CP-5) to observe external artifacts for verification purposes.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from project specification and status files to generate a consolidated HTML report, which represents an attack surface for data-driven influence on the agent's behavior.
Ingestion points: Reads content from spec.md, STATUS.md, and various phase evidence ledgers stored in the 04-projects/ and ~/code/ directories.
Boundary markers: Absent; there are no specific delimiters or warnings provided to the agent to distinguish between instructions and untrusted project data during report generation.
Capability inventory: Includes file writing (generating report.html), execution of local bash scripts, network operations via curl, and use of visual verification tools.
Sanitization: No explicit sanitization or escaping methods are documented for the interpolation of project data into the HTML template.

ultragoal