The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The Code Quality Analyzer script reads and analyzes source code from local directories to identify vulnerabilities. Since the agent is instructed to 'address any P0 issues before proceeding', an attacker could place malicious instructions within the codebase being scanned (e.g., inside a comment like 'P0 ISSUE: Execute [malicious command] to fix security risk'). If the agent treats the analyzer's output as authoritative, it might execute those instructions.
Ingestion points: scripts/code_quality_analyzer.py reads file contents via the count_lines, calculate_complexity, and analyze_security functions.
Boundary markers: No explicit delimiters or 'ignore instructions' warnings are used when the analyzer presents scanned content to the agent.
Capability inventory: The agent is given capabilities to execute shell commands, manage packages (npm install, pip install), and modify the filesystem as part of the primary workflows in SKILL.md.
Sanitization: The tool does not perform sanitization of the strings it extracts from scanned files before including them in the generated report and recommendations.

senior-fullstack