Agent Architect
v1.0.0Audit and improve an agent at the right layer: persona/tone, constitutional and operating rules, memory architecture, or skill portfolio / reusable workflows...
⭐ 0· 86·0 current·0 all-time
byDon Zurbrick@zurbrick
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name and description match the contents: an architecture-focused audit skill. It requests no binaries, no credentials, and contains only documentation and reference files — all proportionate to an auditing/diagnostic task.
Instruction Scope
The SKILL.md stays within scope: it asks the agent to capture symptoms, consult the included reference docs, pick a lane, and produce a small, targeted patch recommendation. It also recommends logging architecture recommendations to daily memory when a structural change is justified — this is expected for an audit skill but is a data-persistence action the operator should be aware of.
Install Mechanism
No install spec and no code files — instruction-only skill. Nothing is written to disk or downloaded at install time.
Credentials
The skill declares no required environment variables, credentials, or config paths. The references and runtime instructions do not request external secrets or unrelated credentials.
Persistence & Privilege
always:false (normal) and model invocation not disabled (normal). The only persistence action is an explicit recommendation to log certain audit entries to daily memory; this is reasonable for an audit tool but means the skill may persist summary notes into the agent's memory store if allowed by the platform — operators should confirm that is acceptable.
Assessment
This skill is coherent and low-risk: it is instruction-only, asks for no secrets, and stays within its stated audit purpose. Before enabling it, confirm you are comfortable with the skill writing brief audit summaries to the agent's daily memory (the SKILL.md recommends doing so for real structural recommendations). Also follow the skill's own advice: review any suggested edits before applying them, and avoid applying broad constitutional changes based on a single run.Like a lobster shell, security has layers — review code before you run it.
latestvk97dmztsqvdddrppkrzneaqmen8382z3
License
MIT-0
Free to use, modify, and redistribute. No attribution required.