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 commands/aleksandarbisevac/claude-plugins/initgit clone --depth 1 https://github.com/AleksandarBisevac/claude-pluginsWrote 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/commands/aleksandarbisevac/claude-plugins/init)<a href="https://agentmods.dev/commands/aleksandarbisevac/claude-plugins/init"><img src="https://agentmods.dev/badge/commands/aleksandarbisevac/claude-plugins/init.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 | $0.00092 | $0.06152 |
| Opus 5 | $0.00046 | $0.03076 |
| Sonnet 5 | $0.00018 | $0.01230 |
| Haiku 4.5 | $0.00009 | $0.00615 |
Grade A, and why
init 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 4d 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 — 406 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/audit:init — generate the audit manifest
Produces the manifest that the /audit:* execution commands run. Generation is multi-agent: parallel
read-only explorers audit the codebase, the orchestrator (you) synthesizes their
findings into phases/tasks. $ARGUMENTS (free text) seeds the interview answers.
0. Conventions
Read ${CLAUDE_PLUGIN_ROOT}/reference/manifest-conventions.md FIRST. Resolve
manifestPath from .claude/audit.config.json (default docs/audit/audit-plan.json).
1. Preflight
If a file already exists at manifestPath, first print any parked proposals it
carries (proposals[] entries with status: "proposed"):
N parked proposal(s) — /audit:propose materialize <id>|--all — the user may be
re-running init when what they actually want is to materialize what the last run
parked. Then ask (AskUserQuestion):
- Abort (default) — keep the existing manifest; suggest
/audit:status. - Regenerate — back it up to
<manifestPath>.bak-<UTC timestamp>first (the backup includes any proposals). - Append phases — keep existing phases; new phases continue the id sequence,
counting BOTH live phases and any
proposals[].payloadreserved phase ids (see manifest-conventions.md → ID allocation). New proposal ids continue thePROP-<n>sequence the same way.
On Regenerate or Append, take the concurrency lock (see
manifest-conventions.md → Concurrency lock) BEFORE touching the file: refuse
while another session holds the index lock, so a generation never clobbers
an in-flight run, and hold it through write + validate (released in step 8).
2. Interview (BEFORE any exploration)
Merge $ARGUMENTS with answers to (ask only what $ARGUMENTS doesn't cover):
-
Dimensions (multi-select, and ALL SIX are offered): security · correctness · test coverage · performance · architecture · DX/build health.
Ask them as TWO questions inside ONE
AskUserQuestioncall — that tool takes at most four options per question, and six do not fit. This is not cosmetic. Asked as one question the last two are cut, and the only route left to them is typing their exact name into the automatic "Other" field, so a reader who has never opened this file cannot know they exist. That is the silent cap this command forbids for areas further down — "a silent cap would read as 'that is all of them'" — applied to its own list first. -
Scope: directories to include/exclude (default: whole repo minus vendored/generated code).
-
Development branch (default
main) →meta.developmentBranch. -
Known pain points — free text. Never a list you synthesize.
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.
- 4d ago First seen · 406 lines · 92 tokens per session scan A 8a428ecea6c1
init is a command published in the GitHub repository AleksandarBisevac/claude-plugins (4 stars, last pushed 3d ago), licensed MIT. It adds 92 tokens to every session and 6,152 once invoked, about $0.0005 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 commands, from other repositories
init
在既有專案初始化 Octopus——摸專案、檢查 OpenSpec CLI 與結構（缺則引導安裝）、盤點 change 狀態、建 Arena、出交接報告。每個專案跑一次。.
list
Show the project backlog (open items by default, sorted by priority then age). Ask for --all, --sort date or --priority high and Claude re-runs the engine with them.
build
SDD 後半段（全自主）：驗 Locked → Builder 逐 task 實作（每條 task 隨行回報）→ 審查 → P1 自動退修（≤3 輪）→ 帶驗收報告回來，你只剩 merge 與 archive。中途不停下來等人。.
prune
Delete old closed entries (status done/cancelled) left over from earlier plugin versions. Shows the list first; the deletion itself asks for your approval.
spec
SDD 前半段：需求釐清（反問+挑戰）→ OpenSpec change 一次落檔（proposal＋delta＋tasks，Draft）→ TPM 拍板鎖定（回一個 OK 即可）。鎖定後 build 全自主執行。.
add
Add a task to the backlog manually. Use when the user explicitly asks to record something in the backlog.