The Agent Skills Directory

[METADATA_POISONING]: The .claude-plugin/plugin.json file contains inconsistent and poisoned metadata that does not align with the provided author context.
The author field lists 'Jeremy Dawes' and the repository points to 'https://github.com/jezweb/claude-skills', both of which contradict the skill author identity 'ma1orek'.
The description field is poisoned with raw code snippets and ends abruptly, which can mislead automated processing of the skill metadata.
[INDIRECT_PROMPT_INJECTION]: The skill provides templates for consumer Workers that ingest data from external queues, creating a vulnerability surface where untrusted input could influence agent behavior.
Ingestion points: The message.body object in templates such as templates/queues-consumer-basic.ts, templates/queues-consumer-explicit-ack.ts, and templates/queues-retry-with-delay.ts acts as the entry point for external data.
Boundary markers: The provided templates do not implement delimiters or instructions to ignore embedded commands within the message bodies.
Capability inventory: The templates are designed to perform sensitive operations including network requests via fetch, database writes via D1Database, and storage access via KVNamespace and R2Bucket.
Sanitization: No input validation, escaping, or schema enforcement is applied to the message content before it is processed or used in downstream operations.

cloudflare-queues