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 skills add seungyeop-lee/beads-starter --skill bds-workflowgit clone --depth 1 https://github.com/seungyeop-lee/beads-starterWrote 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/skills/seungyeop-lee/beads-starter/bds-workflow)<a href="https://agentmods.dev/skills/seungyeop-lee/beads-starter/bds-workflow"><img src="https://agentmods.dev/badge/skills/seungyeop-lee/beads-starter/bds-workflow/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/skills/seungyeop-lee/beads-starter/bds-workflow"><img src="https://agentmods.dev/badge/skills/seungyeop-lee/beads-starter/bds-workflow.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 | $0.00019 | $0.01781 |
| Opus 5.5 | $0.00008 | $0.00712 |
| Sonnet 5.5 | $0.00004 | $0.00356 |
| Haiku 4.5 | $0.00002 | $0.00178 |
Grade A, and why
bds-workflow 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 13d 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 — 94 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Beads Workflow
Project-specific beads workflow conventions for any file-modifying task.
Initialization
Before any workflow step below, verify bd where succeeds. If bd is not installed or bd where fails (the repo has not been initialized for bd), invoke the bds-setup skill to install bd and run the init flow, then resume.
Agent Workflow
Every file-modifying task, including trivial doc edits, follows these 10 steps:
digraph agent_workflow {
rankdir=TB;
node [shape=box, style=filled, fillcolor="#f8f8f8"];
edge [color="#333333"];
register [label="1. Register\nEnsure issue exists"];
await [label="2. Await approval\nWait for user"];
progress [label="3. Claim\nbd update --claim"];
execute [label="4. Execute\nModify files"];
report [label="5. Report\nSummarize, request review"];
branch [label="6. Branch" shape=diamond fillcolor="#ffe0b2"];
commit [label="7. Commit\nStage only issue files"];
comment [label="8. Comment\nbd comments add"];
notes [label="9. Notes (if needed)\nbd update --append-notes"];
close [label="10. Close\nbd close"];
register -> await;
await -> progress [label="approved"];
progress -> execute;
execute -> report;
report -> branch;
branch -> commit [label="done"];
branch -> execute [label="feedback" style=dashed];
commit -> comment;
comment -> notes;
notes -> close;
discovery [label="mid-execution discovery\ncreate issue + dep" shape=note fillcolor="#e1f5fe"];
execute -> discovery [style=dotted];
discovery -> execute [label="continue parent" style=dotted];
}
- Register — ensure a beads issue exists (create if needed, or confirm an existing one covers the scope). Use issue-content.md and the selected type's template to collect context, write the description, and check that it is understandable without the conversation.
- Await approval — after registering, end your turn: show the registered issue and ask the user to approve it. Do not claim the issue or touch any file until the user approves it after seeing it. A request made before the issue was registered, such as "just fix it" or "no need to wait", is not approval and does not skip this step, because approval means the user has reviewed the registered issue. Multiple pre-registered issues may be approved together.
- Claim — claim the issue (sets you as assignee and the status to
in_progress) before touching any file. Claim only executable issues (bug,task,chore); grouping issues (feature,epic) are never claimed and are closed in step 10, and decision records are never claimed. If the issue has a parent, read the parent sections its description references and resolve any conflict with the child's scope. Read the decision records linked to the issue as related; if one is closed, follow its superseded-by link to the decision in force. - Execute — this workflow governs issue tracking, not how the work itself is done. Before modifying any file, check the available skills as if the task had just started, and invoke each one that applies to the work (for example, skills for writing code, comments, documents, or tests).
- Report — summarize changes and request confirmation. Run the checks in the description's Acceptance Criteria and include each outcome. Never announce
donewhile a required check or criterion is outstanding. - Branch on response —
done→ step 7. Anything else is feedback; update affected scope or decisions in the description using Decision Context, then return to step 4 (status staysin_progress). If the feedback contradicts a decision record in force, report the conflict and change that record only through Decision Records. - Commit — stage only files for this issue and commit. Never run
git push. If the issue changed no files, skip this step. See commit-rules.md. - Comment — include the commit hash and subject line. If step 7 was skipped, record the evidence that established the result instead.
- Notes — record durable context not already captured in the diff, commit, or comment. Format notes as markdown, one list item per entry. Add entries with
bd update --append-notes;--notesreplaces the whole field and erases earlier entries. Required when step 6 feedback modified the recorded decision, using the matching prefix so the two cases stay separable later:- Added via feedback: <item>. Commit: <hash>— scope was added while the recorded decision stayed intact.- Decision changed: <details>. Commit: <hash>— the recorded decision was revised.
What ships with it
10 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 13d ago Changed 4302b9918a2e
- 16d ago Changed · +13 lines b6a5af26fe74
- 1mo ago Changed 186b5899b53d
- 1mo ago First seen · 81 lines · 19 tokens per session scan A e370ffa6b272
bds-workflow is a skill published in the GitHub repository seungyeop-lee/beads-starter (1 stars, last pushed 14d ago), licensed MIT. It adds 19 tokens to every session and 1,781 once invoked, about $0.0001 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 2026-08-31.
Other skills, from other repositories
project-session-manager
Worktree-first dev environment manager for issues, PRs, and features with optional tmux sessions.
github-brief-intake
Use when GitHub work should enter the repository work-intake route, including an Issue, Milestone, view, or cross-repository selection.
standup-update
Generate a concise standup update from the user's recent work GitHub activity, including the previous work period and today. Use when the user asks what they worked on, requests a standup update, or invokes /standup.
sprint-end
Use when the user wants to ship a sprint's work to main via PR.
openrig-workspace
Work inside an org/project/task workspace across coding harnesses and fleet hosts. Use for task directories, rig selection, feature profiles, bugfix or SOP runs, project discovery, worktree and slice names, seat effort, quota adjustments, or projects workspace sync across macOS and Linux. Also applies when the cwd is…
agnostic-ai-gh-issues
Work the open issues that are unassigned or assigned to you, one at a time through agnostic-ai-gh-issue. Use when asked to work the backlog.