The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes a logic for triaging failure reports from external tools such as Miri, ASan, and TSan. This creates a vulnerability surface where a user-provided log (untrusted data) could contain instructions designed to manipulate the agent's reasoning. \n
Ingestion points: User-provided Miri and ASan error messages/logs processed via the triage tables in SKILL.md (Section 3) and miri-ub-patterns.md.\n
Boundary markers: None present; the skill lacks specific instructions for the agent to delimit or ignore embedded instructions within user-provided logs.\n
Capability inventory: Writing Rust code, creating FFI stubs, generating and executing shell commands (cargo, rustup, adb) as described in SKILL.md and platform-sanitizers.md.\n
Sanitization: No sanitization or validation of the input logs is performed prior to processing.\n- [EXTERNAL_DOWNLOADS]: The skill instructs the user/agent to install Rust toolchains and components using rustup and cargo. These operations fetch data from the official Rust project repositories and registries, which are trusted sources for language tooling. Evidence found in SKILL.md (Section 4 and 5) and miri-flags-and-ci.md.\n- [COMMAND_EXECUTION]: Provides standard shell command templates for running Rust tests with specific flags and environment variables to enable runtime validation. Examples include cargo +nightly test with RUSTFLAGS="-Zsanitizer=address" and cargo +nightly miri test. These are legitimate operations for the skill's primary purpose of memory safety analysis.

rust-sanitizers-miri