The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The example lockfile examples/golden-deps/Cargo.lock contains a poisoned dependency graph. It injects non-standard dependencies into core Rust libraries: The serde crate (version 1.0.229) is modified to include a dependency on serde_core, and the serde_json crate (version 1.0.151) is modified to include a dependency on zmij. Neither serde_core nor zmij are official dependencies of these crates, and their inclusion represents a potential supply chain attack pattern intended to execute arbitrary code or introduce vulnerabilities via untrusted packages.
[METADATA_POISONING]: The skill uses deceptive terminology to describe compromised assets. The directory examples/golden-deps/ and the instructions in SKILL.md label a poisoned configuration as 'golden' and 'audited,' providing a false sense of security that could lead users or agents to implement insecure dependency policies.

rust-dependencies