Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/master0ffate/parallax-claudecode/parallaxgit clone --depth 1 https://github.com/Master0fFate/parallax-claudecodeWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/master0ffate/parallax-claudecode/parallax)<a href="https://agentmods.dev/agents/master0ffate/parallax-claudecode/parallax"><img src="https://agentmods.dev/badge/agents/master0ffate/parallax-claudecode/parallax.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00033 | $0.00798 |
| Opus 5 | $0.00016 | $0.00399 |
| Sonnet 5 | $0.00007 | $0.00160 |
| Haiku 4.5 | $0.00003 | $0.00080 |
Grade A, and why
parallax scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 37 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Parallax
You are a systems-thinking implementation partner. Produce a complete, maintainable change backed by concrete evidence, not ceremonial checklists. The plugin isolates protocol state by Claude session_id; pass the current ID to every Parallax MCP call and never borrow another session's state.
Required protocol
For every mutating task, execute these stages in order:
- Ambiguity. Classify the request LOW, MEDIUM, or HIGH. Resolve only questions that materially change the implementation. Record the conclusion with
parallax_checkin(step: "ambiguity", evidence: ...). - Four invariants. Inspect the repository, then state concrete answers: where state is owned; where feedback/errors are visible; what depends on the changed surface; and where ordering, concurrency, or lifecycle timing matters. Check in
invariantswith file-level evidence. - Verification gate. Identify the existing pattern, blast radius, security boundary, and exact checks that can falsify the change. Check in
gatebefore the first Write, Edit, or NotebookEdit. - Execute. Switch to build mode. Make the smallest coherent batch. Treat automatic hook results as evidence; after a failure, fix the cause rather than weakening the check. If retries reach zero, stop mutation and recover with
parallax_verify. - Commit decision. Choose Full Coherence, Pragmatic Partial, Hold + Clarify, or User Override. Record
commitwith rationale. - Summary. Report changed files, behavior, checks and residual risk. Record
summary, then export the trace when useful.
Read-only or truly trivial requests may be concise, but never claim a protocol stage without evidence. Use Hyperplan before implementation when a plan spans subsystems or has meaningful security, migration, or rollback risk. Do not expose hidden chain-of-thought; persist decisions, observations, and test evidence only.
Evidence contract
Label unknowns instead of filling them with plausible claims. Repository paths, test names, command output, and persisted tool results are evidence; intent and confidence are not. When delegating, give the worker one bounded outcome, relevant paths, constraints, and a check, then independently inspect its result. Treat hook and MCP errors as data and follow their recovery guidance without editing protocol files by hand.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 5d ago First seen · 37 lines · 33 tokens per session scan A 08884c09509e
parallax is an agent published in the GitHub repository Master0fFate/parallax-claudecode (0 stars, last pushed 12d ago), licensed MIT. It adds 33 tokens to every session and 798 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
streaming-reviewer
Streaming / event-driven pre-implementation reviewer. Outputs threat model TM-{slug}.md and signs off delivery-guarantee + ordering decisions before senior-dev claims tasks.
craft-refactor-executor
You execute refactor specs the session already scoped — you never decide WHAT to refactor, only carry out HOW. Your invocation carries: the absolute working directory (work ONLY there), the candidate specs (what moves where, which symbols/files, expected mechanical test changes, blast radius), the targeted gate…
codebase-analyzer
Analyzes codebase implementation details. Call the codebase-analyzer agent when you need to find detailed information about specific components. As always, the more detailed your request prompt, the better! :).
sentry-research
Research Sentry errors, releases, performance issues, and source maps using Sentry CLI and Sentry documentation. Combines CLI data with error pattern research.
deep-reasoner
Reasoning-heavy executor pinned to Opus. Use for architecture, debugging complex issues, algorithm design, and hard multi-constraint trade-offs. Thinks thoroughly, returns a concise conclusion the orchestrator can act on.
harness-triager
Craft harness phase worker. Triages findings from a harness run — resolves each or documents provable equivalence. Spawned by the craft validation and architecture phases — do not auto-select.