The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill uses dynamic context commands to check the local environment for the nmap binary and helper scripts when the skill is loaded. These commands are executed silently to inform the agent of the system state.
Evidence in SKILL.md: !command -v nmap > /dev/null 2>&1 && echo "YES -- $(nmap --version 2>/dev/null | head -1)" || echo "NO -- Install: brew install nmap (macOS) | apt install nmap (Debian/Ubuntu)"
Evidence in SKILL.md: !test -f scripts/nmap/identify-ports.sh && echo "YES -- use wrapper scripts for structured JSON output" || echo "NO -- using standalone mode with direct commands"
The commands are informational, checking tool presence and version without performing network operations or accessing sensitive files.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided network targets for scanning, which creates a potential surface for indirect injection if scanned services return malicious payloads in banners or response headers.
Ingestion points: The <target> argument used across all nmap and script commands in SKILL.md.
Boundary markers: Not explicitly defined for the processing of scanner output.
Capability inventory: Shell command execution (nmap, bash) and network communication.
Sanitization: The skill includes a target validation mechanism that checks targets against .pentest/scope.json via a PreToolUse hook to ensure scanning remains within authorized boundaries.

nmap