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.
git clone --depth 1 https://github.com/theam/claude-dev-kitWrote 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/theam/claude-dev-kit/plan-backlog)<a href="https://agentmods.dev/commands/theam/claude-dev-kit/plan-backlog"><img src="https://agentmods.dev/badge/commands/theam/claude-dev-kit/plan-backlog/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/theam/claude-dev-kit/plan-backlog"><img src="https://agentmods.dev/badge/commands/theam/claude-dev-kit/plan-backlog.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00057 | $0.00852 |
| Opus 5.5 | $0.00023 | $0.00341 |
| Sonnet 5.5 | $0.00011 | $0.00170 |
| Haiku 4.5 | $0.00006 | $0.00085 |
Grade A, and why
plan-backlog 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 1mo 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 — 40 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Turn the idea/brief given in the arguments into a backlog: $ARGUMENTS
Follow the plan-backlog skill. It has two modes; pick per invocation (never persist a mode — a mature project can hold a brand-new feature):
Rules:
- If the arguments contain no idea or source, ask for one — do not invent scope. Accept pasted text, a path to a file (PDF / Word / Markdown), or a URL / artifact.
- The tracker is whatever
.claude/dev-kit.jsonconfigures (Jira / Linear / GitHub Issues / Azure DevOps); if it is unconfigured, rundev-kit-setupfirst.
Default — guided (run it in THIS conversation)
The guided flow is interactive, so conduct it in the main conversation (do not hand it to a context-isolated subagent — the user needs to make a decision at each level):
- Intake + discovery — read the source; learn the tracker's conventions (issue types, fields, a sample of recent issues). You may delegate the tracker discovery to the
backlog-plannersubagent to keep this context clean. - Three zoom levels, per the skill — at each, present 2–4 alternatives, recommend one with a reason, and wait for the user's decision before going deeper:
- Framing (zoom-out): the goal, the problem space, and framing options (MVP vs full, ways to slice it).
- Epics / themes for the chosen framing.
- Stories (with Given/When/Then acceptance criteria) within each epic.
- Approval gate: print a concise hierarchy summary in the conversation and ask the user to approve, adjust, or cancel. For a non-trivial backlog, also render the full draft as a navigable artifact (collapsible epics → stories, searchable) as the rich review surface — but the approval still happens in the chat (the artifact is not a substitute for the gate). See the
plan-backlogskill's review step. (The per-level decisions do not replace this final gate.) - Create on approval, via the tracker adapter (you may delegate the creation writes to
backlog-planner), then report each item's key/URL and the handoff note (each story ready for/work-story <KEY>).
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.
- 1mo ago Changed · +14 lines · +16 tokens per session a36c2fd79579
- 1mo ago First seen · 26 lines · 41 tokens per session scan A 366e789bee8a
plan-backlog is a command published in the GitHub repository theam/claude-dev-kit (15 stars, last pushed 28d ago), licensed Apache-2.0. It adds 57 tokens to every session and 852 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-09-04.
Other commands, from other repositories
poc
Before building on an unproven idea, get a timeboxed yes/no answer — start a hypothesis-driven POC that skips 80% of the production pipeline and forces ship/pivot/kill at expiry, then promote a shipped POC through the audits it skipped.
spec
Know the problem but not yet the product, the requirements or the design? Run it — you get, stage by stage, a discovery plan (/spec discover), a PRD (/spec prd) and a build spec: requirements.md + design.md + tasks.md (/spec or /spec build). Run before writing any code.
ownership
When you need to know who owns a path and who answers for it — the service ownership matrix (team, lead, on-call, SLA) auto-detected from git with CODEOWNERS generated, plus on-call rotations, shift handoffs and escalation paths.
rfc
RFC process for cross-team decisions. Create, track, and close RFCs. Accepted RFCs auto-create ADRs.
exception
Signed gate-exception registry — replace ad-hoc --admin / --no-verify bypasses with an auditable, signed, expiring record (who · why · gate · scope · expiry). Create / list / check exceptions.
backlog
The project layer: milestones, epics, and user stories with refusing controllers — spec-seeded, sprint-ready.