The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands (curl and echo) to perform local notifications and log execution history to ~/.claude/PAI/MEMORY/SKILLS/execution.jsonl. These commands interpolate user-supplied data such as workflow names and input summaries. There is a risk of command injection if the agent does not properly escape characters like backticks or single quotes when populating these placeholders.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted user input (the problem description) and processes it through workflows that eventually trigger shell command execution.
Ingestion points: User input is ingested in the main SKILL.md workflow for the purpose of generating an '8-word summary' for logging.
Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the user-provided data before it is passed to the shell.
Capability inventory: The skill has the capability to execute shell commands via curl and echo as documented in SKILL.md and the individual workflow files.
Sanitization: There is no explicit sanitization or escaping logic defined for handling the user-provided strings before they are interpolated into the shell execution strings.

SystemsThinking