The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external, untrusted sources including bug trackers and error streams, creating a surface for potential prompt injection attacks via the ingested content.
Ingestion points: External API endpoints and file globs defined in topology.routing.watch.sources and telemetry_sources.
Boundary markers: The instructions do not specify the use of clear delimiters or instructions to ignore potential commands within the ingested data during the classification step.
Capability inventory: The skill has the authority to update product-context.yaml and commit changes to the repository, which directly influences the behavior of downstream automated agents.
Sanitization: There is no documented logic for sanitizing or escaping the content fetched from external sources before it is interpreted or stored.
[PERSISTENCE]: The instructions describe how to implement a continuous 'watch loop' by configuring system-level scheduling utilities such as cron jobs or macOS launchd services to ensure the skill executes at regular intervals.
[COMMAND_EXECUTION]: The skill executes various shell commands for workspace management and validation, including git for repository synchronization and npx for running YAML validation utilities (js-yaml, tsx).
[EXTERNAL_DOWNLOADS]: The skill performs automated network requests using curl to retrieve telemetry findings from remote endpoints as part of its primary ingestion workflow.

aep-watch