The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the installation of several third-party Python packages from public registries, including git-dumper, GitHacker, svn-extractor, and ds-store using pip. These tools are not from recognized trusted sources and are executed at runtime to process remote content.
[COMMAND_EXECUTION]: The methodology utilizes numerous shell commands such as curl for target probing, git for repository auditing, sqlite3 for database querying, and grep for credential searching. These commands are used to interact with remote server responses and local file structures.
[DATA_EXFILTRATION]: The skill explicitly directs the search for highly sensitive files, including SSH private keys (~/.ssh/id_rsa), environment configurations (.env), and database backups (backup.sql). While presented as a methodology for auditing targets, these instructions could lead to the exposure of the host environment's sensitive data if not strictly scoped.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze potentially attacker-controlled data from remote targets.
Ingestion points: The agent processes output from curl requests, content from recovered .git and .svn objects, and directory listings from .DS_Store files.
Boundary markers: No boundary markers or instructions to ignore embedded commands in the target data are provided.
Capability inventory: The skill possesses extensive capabilities including network access via curl, file system read/write, and the ability to install and execute third-party code.
Sanitization: There is no evidence of sanitization or validation of the content retrieved from external targets before it is processed by the agent.

information-disclosure-methodology