The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because its primary function is to process untrusted, arbitrary prose (requirements documents) and translate them into executable configuration and shell scripts. Maliciously crafted input could attempt to influence the generated output, such as modifying telemetry endpoints or including unauthorized commands.
Ingestion points: The skill reads requirements from inline text or attached files provided by the user in SKILL.md (Step-by-step 1).
Boundary markers: The instructions do not define explicit delimiters or 'ignore' instructions for the agent when processing the untrusted input document.
Capability inventory: The agent can write various configuration files (otel.yaml, collector-config.yaml) and shell scripts (claim.sh), and is instructed to run local Python commands for validation.
Sanitization: No specific sanitization or filtering logic is prescribed for the input text before processing.
[COMMAND_EXECUTION]: The skill instructs the agent to perform local command execution to verify the integrity of the generated artifacts.
Evidence: SKILL.md recommends running python -c "import yaml; yaml.safe_load(open('<file>'))" as part of the verification workflow.
Context: This is a localized security check that uses yaml.safe_load to prevent code execution during deserialization, which is a recommended safety practice.

netdata-config-from-requirements