The Agent Skills Directory

[OBFUSCATION]: The skill documents and instructs the agent to recommend the use of zero-width character injection (U+200B, etc.) to bypass upstream security filters and keyword-matching systems (specifically described for the Antigravity gateway in SKILL.md). This is an intentional evasion technique designed to hide content from automated scanners.- [COMMAND_EXECUTION]: The skill includes several Python and Shell scripts (onboard-keys.py, probe-models.sh, sync-limits.py, label-models.py) that perform network operations and modify local configuration files. These scripts have the capability to process and store sensitive API keys and modify critical application settings.- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect injection by ingesting and processing external data from potentially untrusted sources to populate local configuration files.
Ingestion points: The onboard-keys.py script reads API keys and metadata from external text files provided by the user.
Boundary markers: The scripts use regex patterns to identify keys but do not employ explicit delimiters or 'ignore embedded instructions' warnings for the data being processed.
Capability inventory: The skill uses network-capable tools (urllib, curl) and file-system write operations (Path.write_text, shutil.copy2) to manage gateway settings.
Sanitization: While the scripts validate key formats using regular expressions, the data is directly interpolated into YAML and JSON configuration files without further escaping or sanitization.

cliproxy-integration