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/addit-digital/addit-harness/dev-flownpx skills add addit-digital/addit-harness --skill dev-flowgit clone --depth 1 https://github.com/addit-digital/addit-harnessWhat 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 | $0.00120 | $0.01869 |
| Opus 5 | $0.00060 | $0.00934 |
| Sonnet 5 | $0.00024 | $0.00374 |
| Haiku 4.5 | $0.00012 | $0.00187 |
Grade A, and why
dev-flow 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 3d 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 — 143 lines — stays where its author put it; the contents beside it link to each section on GitHub.
dev-flow — deterministic SDLC orchestration
Drives a work item through the full loop this setup's rules/engineering-loop.md
already describes by hand: investigate → design ⇄ architect-reviewer (converges or
caps) → you approve the plan → implement → qa-engineer verifies → code-review
⇄ fix (converges or caps) → qa-engineer re-verifies. The design-gate and
review-gate loops run as deterministic Workflow scripts (real loop-until-approved
control flow, not the model remembering to keep looping); the one thing that stays a
plain conversational step is your approval of the plan — nothing skips that gate.
1. Gather the request
Take the work request as given. Don't ask for more detail than you need to resolve the fields below — if it's already clear from the request, resolve it silently.
2. Resolve track and needsUX
trackis exactly one of'backend','frontend','both'— never anything else. Infer it from the request (a UI change →frontend; an API/data change →backend; anything touching both →both). Ask viaAskUserQuestiononly if the request genuinely doesn't say and the target repo doesn't make it obvious (e.g. a backend-only service repo settles it without asking).needsUXis a separate boolean, not a track value — true only when the work needs a fresh UX pass (new flow, new screen, a UI change with no existing design to work from). A bugfix or an already-specced change is usuallyfalse. Infer from the request; ask only if it's a genuine toss-up.
3. Resolve investigate
true if the request's scope is still fuzzy enough that feature-investigator
should scope it before any design work starts; false if it's already a clear,
bounded requirement. This is a judgment call you make yourself — don't ask the user
"should I investigate first," just decide from how well-specified the request is.
4. Resolve repo and secondaryRepo
repodefaults to the current working directory — no need to ask, matching every other skill in this setup (code-review,save-plan) that implicitly operates on "the repo you're in."secondaryRepostays unset unlesstrack === 'both'. In that case, do a quick check of the cwd first — e.g. ago.modor a server-flavoredpackage.jsonwith no sibling frontendpackage.json, or vice versa, suggests the other track lives elsewhere. Only if that check can't confirm both codebases live inrepo, ask viaAskUserQuestionwhere the other track's repo is. Don't ask when the cwd is obviously a full-stack monorepo.
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.
- 3d ago First seen · 143 lines · 120 tokens per session scan A b51c72a05c6d
dev-flow is a skill published in the GitHub repository addit-digital/addit-harness (3 stars, last pushed 10d ago), licensed MIT. It adds 120 tokens to every session and 1,869 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
ccg-annotate
AI-driven annotation workflow for adding structured metadata to code. Annotations are indexed and searchable via FTS.
ccg-namespace
Isolate read-only CCG search, documentation discovery, and analysis by namespace. Use when working across multiple repositories or services, preventing cross-project graph leakage, federating supported reads, or traversing materialized cross-namespace references. Do not use for graph writes or scoped updates; the user…
ccg-docs
Generate Markdown documentation from code graphs, build RAG indexes for AI consumption, and lint documentation quality.
ccg-analyze
Explicit-only deep analysis of algorithms, feature pipelines, and code relationships with CCG impact radius, bounded flow tracing, callers/callees, git-diff risk, affected stored flows, and cross-namespace references. Use only when the user explicitly names the ccg-analyze skill in the current request. Do not invoke…
ccg-build
Explicit-only graph ingestion for building, updating, migrating, and postprocessing code-context-graph graphs from source visible to the CCG runtime. Use only when the user explicitly names the ccg-build skill in the current request. Do not invoke merely because a graph is missing or stale, annotations need…
ccg
Fast read-only code discovery with a bounded code-context-graph search and targeted source verification. Use when an ordinary positive lookup or explanation needs an entry point, recorded intent, known-path inventory, or direct relationship evidence. Do not use for absence, completeness, exhaustive inventory, deep…