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/si-ic/backlog-skill/backlognpx skills add SI-IC/backlog-skill --skill backloggit clone --depth 1 https://github.com/SI-IC/backlog-skillWhat 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 | $0.00122 | $0.01343 |
| Opus 5 | $0.00061 | $0.00672 |
| Sonnet 5 | $0.00024 | $0.00269 |
| Haiku 4.5 | $0.00012 | $0.00134 |
Grade A, and why
backlog scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 98 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Backlog
Single reliable route for deferred tasks. Storage: docs/backlogs/ in the current
project (the engine resolves the git root automatically). The engine owns the
format — always go through it, never hand-edit files.
Engine: python3 "${CLAUDE_PLUGIN_ROOT}/scripts/backlog.py" <command>
When to record — three cases only
You NEVER capture silently on your own. Noticing that something diverges a little from the current task is NOT a trigger. There are exactly three cases where a backlog entry is created; outside them, do not touch the backlog:
- The user tells you to. Explicit instruction to defer/record: "put it in the backlog / note it for later / let's defer this / out of scope for this PR — backlog it". Record right away, no re-confirmation.
- You noticed something while working — WITH the user's consent. During the task you find either (a) a bug unrelated to the current task that cannot or should not be fixed in passing this session, or (b) an out-of-scope side task (refactor, extraction, improvement) beyond what you were asked to do. Do NOT record it yourself. Mention it and ask once: "record it in the backlog?" — record only after the user agrees. If they say no, drop it.
- You are running autonomously and nobody can be asked. The test is not
which tool launched you — it is: can you ask the user right now and get an
answer before this session ends? If yes, this is case 2, so ask. If no —
a headless
claude -prun, a scheduled run, a phase of a multi-session plan runner, a subagent with no interactive user — AND a task genuinely needs a decision you cannot make, record it so the decision is not lost. This is the ONLY path where you write without an explicit user go-ahead, and it exists precisely because no human is here to be asked.
Hypotheticals, passing remarks, and tasks that advance the current goal are not backlog material. A task that belongs to the currently open plan stays in the plan's own tracking (the active TodoWrite/Task list or the plan-doc checkboxes) — never mirror it into the backlog.
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 · 98 lines · 122 tokens per session scan A 34fd8386cc8e
backlog is a skill published in the GitHub repository SI-IC/backlog-skill (1 stars, last pushed 10d ago), licensed MIT. It adds 122 tokens to every session and 1,343 once invoked, about $0.0006 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
prd
Generates a Product Requirements Document from a natural language app description. Asks clarifying questions, researches similar apps, defines scope, stack, architecture, and produces a structured PRD that Archon can decompose into a campaign.
unharness
Safely leave Citadel using the active adoption receipt. Produces a no-write, reviewable plan, preserves a portable archive, removes only exact owned material, and reports modified or externally registered surfaces as retained or unknown. Legacy installs must be imported before exact leave is claimed.
promote-idea
Converts an album idea from IDEAS.md into an actual album project in one step. Use when the user says "promote [idea title]", "turn idea into album", or "start working on [idea]".
collab-pm-report
한국 팀의 주간 비즈니스 리뷰(WBR) 보고서 자동 생성 스킬입니다. 일일 노트·완료 태스크·KPI 데이터를 입력받아 임원/팀 두 버전(격식체/구어체)으로 작성합니다. Notion·Linear·Asana·Slack MCP가 가용하면 자동 데이터 fetch, 없어도 자유 텍스트 입력 fallback. 다음과 같은 요청 시 반드시 이 스킬을 사용하세요: "주간보고 작성", "WBR 준비", "이번 주 한 일 정리" "주간 회의 자료", "위클리 리포트", "주간 업무 보고" "임원 주간보고", "팀 주간보고" [책임 경계] 제품 발견·로드맵/마일스톤 기획은…
Atera Tickets
Atera service desk tickets: ticket fields, statuses, priorities, comments, work hours, and billing duration.
pm-portfolio-v1
Portfolio and roadmap batch intake, turn a pile of asks into a prioritised verdict. Use when someone hands a BATCH and wants a decision: "prioritise this backlog", "score these requests", "batch verdict", "what goes in the next sprint or quarter", "turn this backlog document into a roadmap", "sprint recommendation"…