The Agent Skills Directory

[SAFE]: The skill's primary script, scripts/discover-account.sh, is restricted to read-only AWS CLI calls (e.g., resourcegroupstaggingapi and cloudwatch describe-alarms) and local file writes for report generation within the workspace.
[PROMPT_INJECTION]: The skill ingests external data from AWS (tags and alarm names) which constitutes an indirect prompt injection surface.
Ingestion points: Untrusted data enters the agent context through the outputs of aws resourcegroupstaggingapi and aws cloudwatch describe-alarms processed in scripts/discover-account.sh.
Boundary markers: The skill uses markdown tables and JSON structures to separate metadata from instructions, as seen in assets/report-template.md and the script's output.
Capability inventory: The agent is limited to read-only discovery commands (describe, list, get) as defined in the Automation Boundary of SKILL.md.
Sanitization: Data is filtered and aggregated using jq before being rendered into markdown reports, ensuring structured presentation and preventing raw terminal escape sequences.

env-discovery