The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and execute shell scripts directly from remote servers.
Evidence: references/cli-tools/visual_review.md provides a command to install the plannotator CLI via curl -fsSL https://plannotator.ai/install.sh | bash.
Evidence: references/cli-tools/code-map.md includes a command to install the ripwire tool from Red Hat Emerging Technologies' official GitHub repository via curl -fsSL https://raw.githubusercontent.com/redhat-et/ripwire/main/scripts/install.sh | bash.
[DYNAMIC_EXECUTION]: The skill frequently utilizes node -e to execute dynamically generated JavaScript strings for parsing configuration, updating workflow state in stelow.json, and managing stage transitions.
Evidence: references/cli-tools/read-config.sh and references/cli-tools/read-config.md use node -e with shell variable interpolation to extract configuration values.
Evidence: stages/setup.md and stages/context.md use node -e to programmatically modify the stelow.json state file.
[INDIRECT_PROMPT_INJECTION]: The orchestrator possesses an extensive attack surface for indirect prompt injection due to its ingestion of untrusted data and its high-privilege capabilities.
Ingestion points: The skill reads project items from .stelow/inbox/items.md (which can be populated from the host's inbox surface), fetches external data via web-research.md (including last30days-skill), and processes rendered DOM snapshots via agent_browser.md.
Boundary markers: Standard boundary markers or instructions to ignore embedded commands are generally absent when processing these inputs.
Capability inventory: The agent has access to bash, node, write, edit, and subagent tools, which can be leveraged if an injection occurs.
Sanitization: No explicit sanitization or validation of the content fetched from external web sources or the browser is described.
[EXTERNAL_DOWNLOADS]: The skill promotes the installation of third-party agent skills and CLI tools from non-trusted or unverifiable sources.
Evidence: references/cli-tools/web-research.md suggests installing mvanhorn/last30days-skill@last30days.
Evidence: references/cli-tools/codequality-review.md suggests installing cursor/plugins via npx skills add.
Evidence: references/cli-tools/code-map.md recommends global installation of @ast-grep/cli via npm or brew.

stelow-workflow-orchestrator