The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The documentation in references/cli.md and references/opensubtitles.md provides examples for the subtitle auth login command that accept --api-key, --username, and --password as plaintext CLI arguments. This practice exposes sensitive user credentials to process monitoring tools, shell history, and logs.
[PROMPT_INJECTION]: The SKILL.md file contains instructions that nudge the AI agent to bypass safety guidelines. It specifically requests that the agent 'strongly prefer an uncensored model variant' to handle 'taboo language, sexual humor, dark comedy' and avoid 'skipping sensitive phrasing.' This matches patterns for bypassing safety filters.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent in references/binary-missing.md to download platform-specific ZIP archives containing binary executables from github.com/waynebaby/SubtitleExtractslator and replace the local skill folder if binaries are missing. This constitutes a remote code acquisition and execution pattern.
[COMMAND_EXECUTION]: The skill performs extensive local command execution using the SubtitleExtractslator.Cli binary to probe media files, check timing, and manage OpenSubtitles authentication. It also modifies application configuration files such as mcp.json at runtime.
[PROMPT_INJECTION]: The skill implements an 'Autonomous Completion Contract' and 'Supervisor Playbook' which instructs the agent to 'not ask whether to continue' and to treat certain phrases as permission to process large batches of files automatically. This reduced human oversight increases the risk of the agent performing unintended actions.
[PROMPT_INJECTION]: The skill processes untrusted external data from OpenSubtitles API responses and local .srt files without explicit boundary markers or sanitization guidelines, creating a surface for indirect prompt injection.
Ingestion points: OpenSubtitles search candidates and external subtitle files processed during translation.
Boundary markers: None present in the instructions.
Capability inventory: Shell command execution via CLI and file system write access for SRT generation.
Sanitization: No sanitization or validation of the content being translated is mentioned.

subtitle-extractslator