iflow-auto

An unattended workflow for advancing a confirmed epic through implementation stages and adversarial reviews. An epic is a large piece of work made up of smaller stages, while adversarial review actively looks for problems.

In plain words
What is it for?
It runs a selected epic stage, reviews the result, re-queues work when needed, reports status, or previews what would happen in a dry run.
Why use it?
It automates repeated implementation and review cycles while keeping durable progress records and a limit on how many review loops can run. It can stop when the work needs human attention.

Skill for Claude CodeCodexCursor

Install

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.

agentmods
npx agentmods add skills/jepegit/issue-flow/iflow-auto
Any agent
npx skills add jepegit/issue-flow --skill iflow-auto
Clone the repo
git clone --depth 1 https://github.com/jepegit/issue-flow

Made for: Claude Code, Codex, Cursor.

Per session 37 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,483 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00037 $0.02483
Opus 5 $0.00018 $0.01241
Sonnet 5 $0.00007 $0.00497
Haiku 4.5 $0.00004 $0.00248

Measured 2d ago against content hash 5208ec9c5435, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

iflow-auto 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.

.cursor/skills/iflow-auto/SKILL.md · 181 lines

How it starts

The opening of the file, as written. The whole thing — 181 lines — stays where its author put it; the contents beside it link to each section on GitHub.

issue-flow — advanced auto (/iflow-auto)

Follow this skill to run an unattended large-change flow over a confirmed epic: select a stage, drive it through /iflow-cycle, record durable state in auto_status.md, run adversarial inter-epoch review (review), then honour the adversarial loop budget (re-queue or stop-and-ask).

Contract: .issueflows/04-designs-and-guides/advanced-auto-mode.md (criteria table, budget, outcomes).

Input

  • <N> — epic anchor issue number (requires epic<N>_plan.md with Status: confirmed).
  • stage <k> — optional stage index; default = earliest unfinished published stage (issue-flow agent epic-status <N> --jsoncurrent_stage).
  • loops:<n> — override adversarial loop budget for this run (baked default 2 from [issueflow].auto_adversarial_loops).
  • review — run only the adversarial procedure for epic <N> (and optional stage <k>); skip cycle unless a full auto run is also intended.
  • status — print auto_status.md / epic-status and stop (no confirm).
  • dry-run — resolve stage + queue, show what would run, stop (no confirm).

Invoke: type iflow auto in chat, or /iflow-auto from the slash menu (iflow-auto also works).

MODEL & EXECUTION DIRECTIVE

Profile: reasoning — Prioritize deep thinking and careful trade-offs over speed or token economy.

In Cursor: switch to a thinking-capable model before invoking this step (not Auto-only).

Keep scope tight to what this step requires.

Resolve project root (multi-root workspaces)

Before any git, gh, or .issueflows/ path operation in this workflow:

Resolution order (stop when unambiguous):

  1. Explicit hints in slash input — root:<path>, repo:<folder-basename> (directory name, e.g. cellpy-core), or repo:owner/name.
  2. CLI fast pathissue-flow agent resolve [-C <start>] [--from-file <active-file>] [--json]. Use the returned project_root and repo; pass -C <project_root> to other issue-flow agent … subcommands. When the answer came from the workspace registry, the payload sets resolved_via_workspace_default: true.
  3. Branch context — exactly one workspace repo whose branch matches ^\d+- → that root.
  4. Single scaffold — exactly one .issueflows/ tree visible in the workspace → that root.
  5. Workspace default — an issueflow-workspace.toml at the workspace root (created with issue-flow workspace init) may name a default member repo; use it when no scaffold matched above. Tell the user the default was used.
  6. Ambiguousstop and ask; never guess between sibling repos.

Read the full file on GitHub · 181 lines

Changes

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.

  1. 2d ago First seen · 181 lines · 37 tokens per session scan A 5208ec9c5435

Subscribe to this mod's changes

iflow-auto is a skill published in the GitHub repository jepegit/issue-flow (4 stars, last pushed 20d ago), licensed MIT. It adds 37 tokens to every session and 2,483 once invoked, about $0.0002 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.