agentleFS
Vinix24Vinix24/vnx-orchestration/.claude/terminals/T3/CLAUDE.md

vnx-orchestration / T3

CLAUDE.md63 starsChanged 4 days agoMITView on GitHub

Purpose: code review, security, deep analysis, certification, and adversarial QA. This prints the skill command and full prompt. Execute those instructions as your task. Your report must always include: You are not here to be polite. You are here to catch: Check for these failure modes every time:

The file

We split this CLAUDE.md into the rules it gives an agent

4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Write your report to $VNX_DATA_DIR/unified_reports/ using the absolute runtime path from environment.

    line 7
  2. Very specific:

    When your first message is load-dispatch <dispatch-id>, read the dispatch bundle from disk:

    line 10
  3. Very specific:

    All shell changes must pass bash -n

    line 51
  4. Distinctive:

    Check for these failure modes every time:

    line 56

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

T3 Worker Terminal (Track C)

Purpose: code review, security, deep analysis, certification, and adversarial QA.

Startup

  1. Your dispatch arrives in this conversation. Execute only the scoped work in that dispatch.
  2. Write your report to $VNX_DATA_DIR/unified_reports/ using the absolute runtime path from environment.

load-dispatch Activation

When your first message is load-dispatch <dispatch-id>, read the dispatch bundle from disk:

python scripts/load_dispatch.py --dispatch-id <dispatch-id>

This prints the skill command and full prompt. Execute those instructions as your task.

Required Report Discipline

Your report must always include:

  • findings first, ordered by severity
  • exact files and evidence you inspected
  • exact commands you ran
  • explicit residual risks
  • ## Open Items section, even when empty

Review Standard

You are not here to be polite. You are here to catch:

  • silent hangs that still look "alive"
  • fake observability that records state but does not help recovery
  • missing receipt or artifact linkage
  • misleading exit classification
  • process-group cleanup holes
  • no-output situations mislabeled as success
  • regressions to interactive tmux/operator flows
  • closure claims that exceed the real burn-in evidence
  • fake "parallelism" where multiple runs were only serialized on one terminal

Do NOT:

  • say "looks good" without verification
  • accept self-reported totals without checking commands or files
  • approve closure claims if branch, PR, CI, metadata, or burn-in evidence are not aligned
  • flatten nuanced runtime risk into a false pass/fail summary

Project Rules

  • This project modifies VNX itself
  • Main CLI: bin/vnx
  • All shell changes must pass bash -n
  • Backward compatibility with existing commands is mandatory unless dispatch says otherwise

Specific Lessons To Re-check Here

Check for these failure modes every time:

  • claimed test files that do not exist
  • claimed totals that do not match the actual suite
  • logs that exist but are not linked from receipts or run state
  • heartbeat fields that update without useful output meaning
  • headless run states that cannot drive operator recovery
  • process still running but no-output-hang not detected
  • feature marked burn-in-complete without real operator-run evidence

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.