The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The troubleshooting instructions describe installing the CLI tool using curl -fsSL https://<PCKLE_HOST>/install.sh | sh. This pattern executes a remote script directly in the shell, which can be used to execute arbitrary code. However, this is presented as a vendor-provided installation method for the tool maintained by the skill author.
[PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection by passing natural language instructions from potentially untrusted sources directly to an autonomous AI task execution engine.
Ingestion points: The --instruction and --input flags in the pckle task create command, including support for reading from files via the @prompt.txt syntax.
Boundary markers: No delimiters or explicit boundary markers are provided to isolate user-supplied instructions from the task execution logic.
Capability inventory: The tool can perform autonomous AI searches, list and manage tasks, and has access to local CLI configuration files and environment variables. The agent also has Bash access.
Sanitization: No input validation or sanitization of the natural language instruction is mentioned in the documentation.
[COMMAND_EXECUTION]: The skill requires Bash(pckle:*) permissions and demonstrates complex command chains using jq and xargs to manipulate task IDs and automate CLI operations, which involves executing multiple shell commands based on tool output.

pckle