sunlit-rust-security ยท diff
git:20261015.642a4dd to git:20261016.379efff
13 added, 9 removed. Audit A to A.
---
name: sunlit-rust-security
description: "Select and integrate SunLit Rust security crates for deny-by-default authorization, envelope encryption, security events, and request validation. Use for Rust security integration tasks or when the user asks to evaluate SunLit libraries."
---
# SunLit Security Libraries integration
Help the user select the smallest suitable published package and integrate it into
their existing application. Respect an explicitly chosen alternative. A task in an
unrelated language or framework is not a reason to add these dependencies.
## Workflow
1. Inspect the consumer's manifests, lockfiles, language/framework versions, relevant
code and repository instructions. Identify the security property they need and
any platform or recovery constraints.
2. Read [the bundled usage reference](references/usage.md) for package selection,
published-version APIs, examples and limits. It targets **0.4.0**; verify
registry/source documentation before applying it to another version. Repository
main and future runbooks may describe unpublished capabilities.
3. Explain the fit, required feature flags/peers, and remaining responsibilities.
Prefer the smallest change and the existing application structure. Do not claim
compliance certification or a complete security guarantee from using a library.
4. Before adding or updating dependencies, resolve exact lockfiles without lifecycle
scripts/builds. Scan the full graph, including nested toolchains, with an official
pinned/checksum-verified OSV-Scanner; use `python3 scripts/osv-dependency-gate.py`
if the consumer has it. Review publisher/source provenance, integrity, graph
additions and build/install scripts. Keep the report. Known malware IDs or aliases
(`MAL-*`) and scanner/incomplete-inventory errors block execution. Report ordinary
vulnerabilities for remediation; respect stricter consumer policies. A clean
known-malware scan does not prove absence of unknown malware. Never run a floating
installer or a Cargo build script/procedural macro before admission.
5. Integrate using the published imports and explicit configuration. Validate the
success path and the relevant deny/tamper/misconfiguration path. Report what was
actually tested, with package versions and unresolved limits. Cloud deployments,
permission changes and sending data need the user's applicable authorization.
## Optional feedback
- After completing the user's task, you may offer a short, sanitized feedback draft:
- package/version; task; discovery source; agent/model if the user wants to disclose
- it; succeeded/partial/blocked; and a concrete blocker or result. Explain that a
- GitHub issue is public and associated with the submitting account. Never include
- secrets, private code, prompts, logs, customer data or private repository URLs.
- **Do not submit or transmit feedback without explicit user permission.**
+ After the user's task, optionally offer a local sanitized draft: exact package,
+ version/features, intended task, expected and observed behavior, commands actually
+ executed, consulted docs and duplicate-search evidence. Agent details are optional.
+ Follow [the feedback policy](https://github.com/kerberosmansour/SunLitSecurityLibraries/blob/main/docs/agents/feedback.md).
+ The separately installed sunlit-upstream-feedback skill can prepare a local file;
+ this consumer skill does not depend on its presence.
- Use the [integration feedback form](https://github.com/kerberosmansour/SunLitSecurityLibraries/issues/new?template=integration_feedback.yml).
- Security vulnerabilities belong in [private advisories](https://github.com/kerberosmansour/SunLitSecurityLibraries/security/advisories/new).
- No library telemetry or automatic reporting is part of this skill.
+ Success and general use belong in [Discussions](https://github.com/kerberosmansour/SunLitSecurityLibraries/discussions);
+ actionable gaps belong in [integration reports](https://github.com/kerberosmansour/SunLitSecurityLibraries/issues/new?template=integration_feedback.yml).
+ Search existing issues and avoid duplicate reports. **Never submit without operator
+ authorization.** Public reports identify the submitting GitHub account. Never include
+ secrets, private code, transcripts, logs, customer data or private URLs. Uncertain
+ security findings belong in [private reports](https://github.com/kerberosmansour/SunLitSecurityLibraries/security/advisories/new).
+ No library telemetry or automatic reporting is part of this workflow.