The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the user to install the CLI tool by fetching a shell script from https://anycap.ai/install.sh and piping it directly to sh. This pattern involves executing remote code without local inspection.
[COMMAND_EXECUTION]: The skill depends on the execution of shell commands, specifically the anycap binary and the jq utility, to perform its core tasks.
[CREDENTIALS_UNSAFE]: The skill manages sensitive authentication data, including API keys and OAuth tokens, which are stored in the system keychain or a configuration file at ~/.anycap/credentials.
[EXTERNAL_DOWNLOADS]: The tool performs external downloads via its anycap download and anycap update commands, fetching binaries and files from the vendor's remote servers.
[DATA_EXFILTRATION]: Multimodal analysis commands (image-understand, video-read) involve uploading local files to the AnyCap API, which transmits local data to a remote cloud service.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks when processing instructions alongside external media. 1. Ingestion points: references/actions.md via --url and --file flags. 2. Boundary markers: None identified. 3. Capability inventory: File system access, network operations, and CLI-based command execution. 4. Sanitization: No evidence of sanitization for processed external content.

anycap-cli