The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/agent.py executes system utilities such as strings, upx, and readelf using subprocess.run to perform static analysis on user-provided binary files. This is a functional requirement for the skill's stated purpose of malware triage.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from untrusted external binaries, which presents an attack surface for indirect prompt injection.\n
Ingestion points: The extract_strings function in scripts/agent.py and string extraction commands in SKILL.md read content directly from suspect ELF binaries.\n
Boundary markers: The skill does not provide specific delimiters or instructions to the agent to treat extracted strings strictly as data rather than instructions.\n
Capability inventory: The skill has the ability to execute shell commands via subprocess.run for analysis tasks.\n
Sanitization: Content extracted from binary strings is not sanitized or filtered for control characters before being returned to the agent context.\n- [METADATA_POISONING]: Documentation in SKILL.md contains sample indicators of compromise (IOCs) associated with Monero cryptominers, causing automated scanners to flag the file for malware signatures (e.g., BitCoinMiner-BY). These indicators are defanged and serve as instructional examples rather than active malicious code.

analyzing-linux-elf-malware