The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill recommends using sudo apt install k6 in references/load-testing-with-k6.md. Commands requiring superuser access pose a risk of unauthorized system modifications and potential host compromise.
[INDIRECT_PROMPT_INJECTION]: The skill processes output from external load testing tools, which presents a vulnerability to instructions embedded in the data being tested.
Ingestion points: The agent is instructed to read and analyze output from k6 and autocannon runs as described in SKILL.md and references/load-testing-with-k6.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish between the load testing metrics and potentially malicious instructions returned by a target API.
Capability inventory: The skill possesses the Bash (npx, vercel, curl), Write, and Edit capabilities, which could be leveraged to perform harmful actions if the agent is manipulated.
Sanitization: The instructions do not define any validation or sanitization steps for the external tool outputs before they are processed by the agent.
[METADATA_POISONING]: There is a manifest inconsistency where the instructions in references/examples.md suggest using kubectl commands, but kubectl is not included in the allowed-tools section of the SKILL.md frontmatter. This indicates the skill is either misconfigured or attempting to use unauthorized tools.

vercel-load-scale