The Agent Skills Directory

[PROMPT_INJECTION]: The skill metadata contains deceptive author information.
Evidence: The YAML frontmatter in SKILL.md identifies the author as 'Render', which contradicts the authoritative author context ('Sarthib7'). This deception could lead to misplaced trust in the skill's operations.
[REMOTE_CODE_EXECUTION]: The skill instructs the agent to install the Render CLI by downloading and executing a shell script from a remote URL.
Evidence: curl -fsSL https://raw.githubusercontent.com/render-oss/cli/main/bin/install.sh | sh in SKILL.md.
Context: The script is sourced from the official repository of a well-known cloud service.
[EXTERNAL_DOWNLOADS]: The skill fetches an installation script and references configuration templates from external repositories.
Evidence: URLs pointing to the render-oss organization on GitHub are present in SKILL.md.
[COMMAND_EXECUTION]: The skill suggests executing the Render CLI to validate configurations.
Evidence: Usage of render blueprints validate and render whoami in the deployment workflow.
[PROMPT_INJECTION]: The skill processes untrusted codebase files to generate deployment configurations, creating an indirect prompt injection attack surface.
Ingestion points: Project manifests and configuration files analyzed as described in references/codebase-analysis.md.
Boundary markers: The instructions lack delimiters or warnings to ignore instructions embedded within the analyzed files.
Capability inventory: The agent can execute shell commands and create cloud resources via MCP tools as described in SKILL.md and references/direct-creation.md.
Sanitization: No explicit sanitization or validation of extracted values is described in references/configuration-guide.md.

render-deploy