Feedback Analyst — Security agent for Claude Code
Reads the audit log, human edits and reviewer eval results and proposes changes to the Harbourlight rules or voice guide.
How to install Feedback Analyst
Installs to ~/.claude/agents/cla1redonald-harbourlight-feedback-analyst.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/cla1redonald/harbourlight/HEAD/agents/feedback-analyst.md -o ~/.claude/agents/cla1redonald-harbourlight-feedback-analyst.md Restart Claude Code, or start a new session, for it to be picked up.
What Feedback Analyst does
name: feedback-analyst description: Reads the audit log, human edits and reviewer eval results and proposes changes to the Harbourlight rules or voice guide. Proposes only; never applies. Read-only. tools: Read, Grep, Glob
You look for patterns in how people overrule or correct the pipeline, and you turn them into specific, small proposals. A person decides whether to accept them, through `/voice-apply`, with a named approver.
Read
data/audit-log.csv, especially `human_decision
Alternatives in Security
- Vuln Verifier — Adjudicates a single candidate security finding produced by the vuln-analyst hunter 107 ★
- Log Analyst — Security log analysis specialist 72 ★
- Gsd Eval Auditor — Retroactive audit of an implemented AI phase's evaluation coverage 57 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Log Reviewer
Audit roku-log usage in a JellyRock BS/BRS file or function. Defaults to AUDIT-ONLY: lists logging gaps withou
Antwork Voice Analyst
Read-only analysis agent for the Antwork audit. Owns the Voice Consistency dimension — whether each account's
Governance Learner
Feedback Distiller. Use this agent at sprint close to audit /memory/ logs, distill systemic patterns from bug-
Mauro Context Verifier
Adversarially verifies Mauro knowledge candidates against code, tests, human decisions, scope, and security ru
API Review Orchestrator
Entry point for a full API design review. Invoke when the user asks to review, audit, assess, or give feedback
Debate Prosecutor
L0.97 Adversarial Verification — reads diff + spec + test results, constructs the strongest-case argument that
Related Skills
Arreflect
AgentRecall consolidation & reflection — periodic triage of recurring corrections; proposes rule changes, neve
Security Watch
One security-watch pass over an EVM / Solidity repo — deps + static analysis + risky-pattern grep + recent adv
Hatch3r Diagnose
Troubleshoot a hatch3r framework issue (setup, config, adapter wiring, drift). Gathers state, delegates root-c