The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill enables the agent to process data from untrusted binaries through analysis tools like pe_analyze.py, creating a surface for potential indirect prompt injection. 1. Ingestion points: Output from tools/pe/pe_analyze.py and tools/disasm/analyze.py (as described in SKILL.md). 2. Boundary markers: None explicitly mentioned in the instructions to delimit tool output. 3. Capability inventory: The agent is authorized to execute shell commands to run the Python-based toolkit and build systems like CMake and MSVC. 4. Sanitization: No specific sanitization or validation of the analysis output is mentioned before the agent uses it to decide subsequent actions.
[EXTERNAL_DOWNLOADS]: The skill requires the user to clone the pcrecomp repository from GitHub and install well-known Python packages for binary analysis. 1. Evidence: Instructions in README.md and references/tool-commands.md mention cloning from github.com/sp00nznet/pcrecomp.git and running pip install for capstone, pefile, and lief.
[COMMAND_EXECUTION]: The core functionality of the skill involves the agent running command-line tools for disassembly, classification, and lifting. 1. Evidence: Numerous commands such as 'python -m tools', 'python tools/pe/pe_analyze.py', and 'cmake -B build' are listed as primary workflows in SKILL.md and references/packaging.md.

pcrecomp