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 skills/huzaifa525/claude-code-optimizer/subagent-devnpx skills add huzaifa525/claude-code-optimizer --skill subagent-devgit clone --depth 1 https://github.com/huzaifa525/claude-code-optimizerWhat 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 | $0.00030 | $0.00867 |
| Opus 5 | $0.00015 | $0.00434 |
| Sonnet 5 | $0.00006 | $0.00173 |
| Haiku 4.5 | $0.00003 | $0.00087 |
Grade A, and why
subagent-dev 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 — 124 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Execute a plan using subagent-driven development for: $ARGUMENTS
Iron Law
The controller coordinates. Subagents implement. Never mix roles.
The controller agent (you) MUST NOT write production code directly. You dispatch subagents for each task, review their work, and coordinate the overall plan.
Architecture
Controller (you)
├── Read the plan
├── For each task:
│ ├── Dispatch Implementer subagent (fresh context)
│ ├── Dispatch Spec Reviewer subagent
│ ├── Dispatch Quality Reviewer subagent
│ └── Handle status (DONE / CONCERNS / BLOCKED)
└── Verify all tests pass at the end
Steps
1. Load the Plan
Read task_plan.md or the user's task list. Extract all tasks with:
- Task description
- Files involved
- Dependencies (what must complete first)
- Acceptance criteria
2. Execute Each Task
For each task (respecting dependency order):
A. Dispatch Implementer
Launch a subagent with Agent tool:
Prompt: "Implement [task description].
Files to modify: [list]
Pattern to follow: [reference file]
Acceptance criteria: [criteria]
Run tests after implementation."
Use model: "sonnet" for mechanical tasks, model: "opus" for complex architecture.
B. Check Implementer Status
The implementer should report one of:
| Status | Meaning | Action |
|---|---|---|
| DONE | Task complete, tests pass | Proceed to review |
| DONE_WITH_CONCERNS | Complete but has questions | Review concerns, decide if acceptable |
| NEEDS_CONTEXT | Missing information to proceed | Provide context, re-dispatch |
| BLOCKED | Cannot complete (dependency, bug) | Log blocker, skip to next task |
C. Two-Stage Review
Stage 1 — Spec Compliance (must pass first):
- Does the implementation match the task description?
- Are all acceptance criteria met?
- Were the correct files modified?
Stage 2 — Code Quality (only after Stage 1 passes):
- Does it follow existing codebase patterns?
- Are there security issues?
- Is the code clean and maintainable?
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 · 124 lines · 30 tokens per session scan A 2b7d64e3467f
subagent-dev is a skill published in the GitHub repository huzaifa525/claude-code-optimizer (9 stars, last pushed 5mo ago), licensed MIT. It adds 30 tokens to every session and 867 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 skills, from other repositories
throughline
Use when the user asks to use Throughline from Codex, continue or restore Throughline memory, export read-only handoff context for a local launcher, prepare a new Codex thread handoff, summarize a captured Codex session, or check whether the Throughline Codex Stop hook captured the current session. Hide long…
critical-review
Use when you want structured feedback on a plan or document before implementation.
save
Save all session progress to status tracking files. Use when you want to checkpoint work mid-session or before ending.
code-review
Use when you want a thorough code review of files, changes, or the entire project before shipping.
implement-batch
Use when you want to implement the next batch of a plan. Handles module implementation, testing, and validation.
audit-orchestrator
Universal Pre-Scan → Analysis → Optimization → Report orchestrator for ANY project type — web apps (Astro/SvelteKit/Next), infrastructure/homelab repos, CLI tools, libraries, backend services, monorepos, data/ML projects, docs. Self-detects project type and runs the matching analysis track. Session state lives in…