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/klausfreiberufler/devflow-mcp/devflow-loopgit clone --depth 1 https://github.com/KlausFreiberufler/devflow-mcpWhat 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 | $0.00011 | $0.02770 |
| Opus 5 | $0.00005 | $0.01385 |
| Sonnet 5 | $0.00002 | $0.00554 |
| Haiku 4.5 | $0.00001 | $0.00277 |
Grade A, and why
devflow-loop 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 yesterday.
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 — 165 lines — stays where its author put it; the contents beside it link to each section on GitHub.
devflow-loop
Run the active flow end-to-end without leaving the session. The Stop-Hook (DF-405) holds you in until the backend's flowGate service confirms every condition is green. Discipline-tokens are emitted explicitly per state, and knowledge-resolution is callable in review (DF-408) so the pending_drafts_resolved gate never traps you in a catch-22.
This command is the orchestrator. It does NOT replace the per-state skills (devflow-planning, devflow-executing, devflow-reviewing, …) — it sequences them.
Foundation it relies on
- DF-405 — Stop-Hook returns exit 2 + precise stderr-reinject if you try to end the session in a non-wait state. This makes the loop self-correcting.
- DF-323 —
pre-flow-update-self-approval.jshook surfaces required-skills automatically on everyflow_updateattempt. - DF-408 —
KNOWLEDGE_RESOLUTION_TOOLS(knowledge_draft_accept/reject, knowledge_check_resolve, intent_resolve) callable inapproval+review. No backward-transition workaround needed. pipelineOrchestrator.HARDCODED_REQUIRED_SKILLS— single source of truth for which discipline-tokens each transition needs.
Argument Resolution
$1is optionalflowId. Resolve:- Starts with
DF-→ use as-is - Purely numeric (e.g.
214) → prefixDF-(DF-214) - Otherwise → pass through (assume full flow id like
1779181904709-l4pfwtyvu) - If missing → read
.devflow-activein the repo root and use itsflowIdfield
- Starts with
--max-iterations N→ integer, default40--verify "<shell-cmd>"→ shell command to run beforereview. Its stdout/stderr is captured intotestingInstructionsfield. If absent, use the project'sverifyscript if it exists inpackage.json, otherwise skip and note intestingInstructionsthat verification was manual.
Iron Laws
- Only use real MCP tools.
devflow_init,flow_get,flow_update,devflow_token_emit,devflow_tokens_list,knowledge_check_flow,knowledge_check_resolve,knowledge_draft_create,knowledge_draft_accept,knowledge_draft_reject,task_create,task_update,task_list,planning_context,wiki_get_briefing,pending_work,flow_upload. Never invent tool names. - No "should work" claims. Every acceptance criterion gets evidence from the
--verifycommand's actual output. Capture stdout verbatim intotestingInstructions. - Never hand-fabricate tokens. All discipline-tokens come from
devflow_token_emit({flowId, skillName, evidence}). The backend signs them; the unsigned token is returned once. - Respect the gate. A 409 response carries
gate.failures[]with named conditions (e.g.plan-required,tasks-required,agent-summary,testing-instructions,adr-compliance-clean,pr-state-consistent). Fix the named field, then retry the SAME transition. Never invent fields the gate didn't mention. - Stop-Hook does the policing. You cannot legitimately end the session in
idea/planning/ready/in_progress. If you think you're done, transition forward — the Stop-Hook will reject premature exit otherwise (DEVFLOW_STOP_HOOK_SOFT=1is the only override and is gated).
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.
- yesterday First seen · 165 lines · 0 tokens per session scan A 1e52f2ffceb8
devflow-loop is a command published in the GitHub repository KlausFreiberufler/devflow-mcp (1 stars, last pushed 3d ago), licensed MIT. It adds 11 tokens to every session and 2,770 once invoked, about $0.0001 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
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.
implement
Execute the implementation plan by processing and executing all tasks defined in tasks.md.