The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill has a standard attack surface for testing-related tools as it processes external, potentially untrusted data.
Ingestion points: The skill processes external API specifications (e.g., openapi.yaml), Pact contract files (e.g., pacts/*.json), and GraphQL schemas (e.g., schema.graphql) as seen in SKILL.md and references/agent-commands.md.
Boundary markers: There are no explicit instructions or delimiters used to isolate or warn the agent about treating the content within these processed files as untrusted.
Capability inventory: The skill utilizes subprocess calls for CLI tools like aqe, npm, and npx, performs file writes for generating contracts, and mentions network operations for publishing to a Pact Broker.
Sanitization: No explicit sanitization or validation logic for the content of the ingested API specifications or contract interactions is described beyond standard schema compliance.
[EXTERNAL_DOWNLOADS]: The skill documentation references the use of external tools and services.
SKILL.md mentions npx pact-broker publish, which may download the Pact Broker CLI from the npm registry at runtime if not already installed.
references/provider-states.md includes an example using curl to interact with a Pact Broker API for webhook configuration.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands using the aqe CLI tool and Node.js package managers (npm, npx) to perform testing and validation tasks.
[SAFE]: The skill follows security best practices for credential management by instructing the use of environment variables (e.g., process.env.PACT_TOKEN) rather than hardcoding secrets.

contract-testing