The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, specifically ransom notes (README/DECRYPT files) and encrypted file headers. This constitutes a vulnerability surface where malicious actors could embed instructions to manipulate the agent's behavior during analysis.
Ingestion points: SKILL.md Step 1 (searching for ransom notes and using file on encrypted targets) and Step 4 (extracting data from ransom notes).
Boundary markers: None specified for the ingestion of note content.
Capability inventory: The skill utilizes subprocess commands (sha256sum, xxd, find) and references script execution via linked skills (re-crypto-decrypt).
Sanitization: No explicit sanitization or filtering of ransom note content is mentioned, although the instructions mandate sandbox isolation for execution.
[COMMAND_EXECUTION]: The skill involves executing shell commands to analyze potentially malicious files, which could lead to exploitation if the underlying tools have vulnerabilities triggered by malformed file structures.
Evidence: Use of sha256sum, file, and xxd in Step 1 against encrypted files and malware samples.
[EXTERNAL_DOWNLOADS]: The skill references established security resources for malware family identification and decryption tools.
Evidence: Mentions of id-ransomware.malwarehunterteam.com and nomoreransom.org in the Tools section. These are well-known and reputable services within the cybersecurity community.

re-ransomware