The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from prd.md, architecture.md, and task-checklist.md to drive its execution. This creates an indirect prompt injection surface where instructions in these documents could influence the agent's behavior during code implementation or command execution.
Ingestion points: docs/juspay/prd.md, docs/juspay/architecture.md, and docs/juspay/task-checklist.md located in the project workspace.
Capability inventory: Shell command execution (npm, gradle, pod, pub), file system writes (source code and .env), and tool calls to the juspay-mcp server.
Boundary markers: The skill does not define explicit delimiters for interpolated content but mitigates risk via "grounding" rules that require re-fetching authoritative documentation.
Sanitization: No explicit sanitization or validation of the input document content is mentioned before processing.
[COMMAND_EXECUTION]: The skill executes various package management and platform-specific commands to install SDKs and configure the project workspace. This includes running post-install scripts and server start commands for smoke testing.
Evidence: npm install, gradle, pod, and pub commands in step-03-core.md, as well as automated configuration scripts in step-06-native-setup.md and server execution in step-08-test.md.
Context: Command execution is a primary function of this skill and is restricted to packages and steps defined in the fetched documentation.
[EXTERNAL_DOWNLOADS]: The skill fetches authoritative documentation from Juspay-controlled domains to guide the integration.
Evidence: Target domains include *.juspay.io, *.juspay.in, and an official Cloudfront distribution (dth95m2xtyv8v.cloudfront.net) accessed via the docs-mcp-server.
Context: These are legitimate vendor resources used for documentation grounding.
[CREDENTIALS_UNSAFE]: The skill manages sensitive API keys and webhook credentials by provisioning them through the juspay-mcp or guiding the user through manual entry.
Evidence: step-02-credentials.md and step-03-core.md handle the generation and storage of credentials.
Context: The skill adheres to security best practices by ensuring secrets are handled in memory and stored only in .env files, with strict prohibitions against logging secrets or including them in public artifacts.

jp-executor