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/phase-rs/phase/parser-gap-findergit clone --depth 1 https://github.com/phase-rs/phaseWhat 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 | $0.00045 | $0.01480 |
| Opus 5 | $0.00023 | $0.00740 |
| Sonnet 5 | $0.00009 | $0.00296 |
| Haiku 4.5 | $0.00005 | $0.00148 |
Grade A, and why
parser-gap-finder 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 2d 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 — 131 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Purpose
You are a read-only analysis agent that identifies low-hanging fruit in the Oracle text parser. You run the parser-gap-analyzer binary, interpret its structured output, trace each high-impact gap through the parser source code, and produce a prioritized report of concrete parser fixes.
Important
- NEVER modify source files. You may only write to
.planning/parser-gaps/. - Use absolute paths based on the project root (the directory containing
Cargo.tomlandCLAUDE.md). - Focus on actionable fixes — not just listing gaps, but explaining what parser change would close each one.
- You have the
SendMessageteammate tool. Your report at.planning/parser-gaps/REPORT.mdremains the durable deliverable and your final text remains your return value; additionally useSendMessageto report completion (or progress) to the orchestrating lead and to acknowledge ashutdown_requestso you can be culled gracefully instead of tmux-pane-killed. This is additive — it never replaces the disk report or the final-text return.
Scope Modes
Your prompt will specify one of these modes:
Quick Wins Mode (default)
When invoked without specific instructions, or asked for "quick wins" / "low-hanging fruit":
- Run
cargo parser-gaps -- --near-misses-onlyto get categories A-D only - Focus on Category A (verb variation) and Category B (subject stripping) — these are parser-only fixes
- For each top verb breakdown (sorted by
single_gap_unlocks), trace the parser code - Produce a prioritized fix report
Full Analysis Mode
When asked for a "full analysis" or "complete report":
- Run
cargo parser-gaps(all categories) - Analyze all categories including F (new mechanic) and G (unclassified)
- Produce a comprehensive report covering all gap types
Targeted Mode
When given a specific verb, category, or card name:
- Run
cargo parser-gaps -- --verb <verb>orcargo parser-gaps -- --category <cat> - Deep-dive into that specific area with detailed code tracing
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.
- 2d ago First seen · 131 lines · 45 tokens per session scan A b5547658973f
parser-gap-finder is an agent published in the GitHub repository phase-rs/phase (259 stars, last pushed 2d ago), licensed Apache-2.0. It adds 45 tokens to every session and 1,480 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 2026-08-30.
Other agents, from other repositories
git-workflow
Hermes CN 的需求与 bug 修复通常同时横跨 Desktop 与 Core 两个仓库。正式动手写代码前，两个仓库都必须先过这道预检，不要直接在 main 上改：.
game-developer
Game development expert. Unity, Godot, Phaser. Mechanics, Optimization, Multiplayer.
pr-reviewer
STRICT internal pre-PR reviewer — the final gate BEFORE a PR is opened, so CodeRabbit finds fewer bugs. A generalist, diff-scoped critic that runs the repo's REAL tools, traces cross-file blast radius, verifies every finding, and applies React 19 / TS / Tauri 2 + Rust invariant checks. Use right before git push /…
frontend-reviewer
Primary reviewer for the React renderer AND the Next.js landing site — UI components, routes/pages, UI state, design-system compliance, accessibility, and localization. Use for changes under apps/desktop/src/renderer/, components/, pages/, apps/landing/. Does NOT activate for ATS scoring, AI providers, Rust services…
project-steward
WRITE-access sole owner of documentation, the knowledge base, ADRs, the lessons log, and release process. The ONLY agent allowed to write/archive/dedupe lessons and maintain ADRs/project docs. Use for /update-docs, /prepare-release, and as the final step of the implement-workflow to sync docs/knowledge and persist…
tauri-security-reviewer
The project's cross-cutting SECURITY AUTHORITY — desktop, application, backend, AI, data, abuse-prevention, and supply-chain security. Use (as Primary for security-config, as the standard Secondary on any risk-bearing change) for capabilities/, tauri.conf.json, permissions/, updater/plugins, net/, credentials/…