Diagnose internal thinking trace leaks in Samara message outputs.
npx skills add https://github.com/claudeaceae/samara-main --skill diagnose-leaks
Please help me install this Agent Skill. Skill: diagnose-leaks Source: https://github.com/claudeaceae/samara-main/tree/main/.claude/skills/diagnose-leaks Command: npx skills add https://github.com/claudeaceae/samara-main --skill diagnose-leaks
This Skill helps identify and fix instances where internal thinking traces, session IDs, or other hidden content leak into user-visible messages, ensuring safer and cleaner interactions.
Run the diagnostician to audit recent sessions and generate a leak report using its built-in scripts and log checks.
High-intent search queries and answers about installing and using this skill.
To diagnose thinking-trace leaks, run the diagnostician to audit recent chat sessions, applying layered sanitization checks and log analysis to generate a report verifying hidden content is properly sanitized.
Thinking trace leaks occur when internal thinking blocks or session identifiers leak into user-visible messages. They typically occur in multi-task chat sessions when MessageBus and TaskRouter layers fail to prevent cross-contamination.
You verify sanitization checks by running scripted diagnostics that inspect message outputs and log pipelines, confirming session identifiers and thinking blocks are properly sanitized and preventing hidden content exposure.
The best way to isolate cross-contamination is using routing and isolation checks that validate MessageBus and TaskRouter layers, ensuring they prevent hidden content from leaking across multi-task chat sessions.
Yes, you can use automated diagnostics to audit chat histories and log pipelines, running built-in scripts and log checks to generate a clear remediation report for thinking trace and session identifier leaks.
DDTullyBuild Streamlit chat UIs with st.chat_message and st.chat_input.
edmundmillerExplain commit visibility in jj log using revset filters.
SnakeODiagnose and fix Next.js SSR, hydration, routing, and build-time issues.
dapperdiversGenerate structured markdown penetration test reports from JSON security scan findings.
