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/surjal/dev-agent/implementgit clone --depth 1 https://github.com/Surjal/dev-agentWhat 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 | $0.00060 | $0.06991 |
| Opus 5 | $0.00030 | $0.03495 |
| Sonnet 5 | $0.00012 | $0.01398 |
| Haiku 4.5 | $0.00006 | $0.00699 |
Grade A, and why
implement 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 — 448 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Run the full development workflow for this task:
$ARGUMENTS
This plugin's subagents are OpenCode markdown agents (.opencode/agents/dev-agent-*.md, mode: subagent). Delegate to one by calling the task tool with its agent name (e.g. dev-agent-developer,
dev-agent-researcher) -- never do the research/design/implementation/testing/review yourself
inline, even when the change looks small. Task size is not a reason to skip delegation -- actually
call the task tool for every stage that applies, every time.
Deep Mode (default -- this is not an opt-in flag)
Every /implement run is Deep Mode. There is no separate "fast mode" -- a small bug fix and a new
application both go through the same 15-stage lifecycle below, they just skip more of it (and skip
explicitly, see Stage selection). --deep is accepted as a no-op for anyone who types it out of
habit; do not treat its absence as license to behave more shallowly.
Execution philosophy
This is an autonomous, quality-first execution system. Do not optimize for speed, token usage, number of tool calls, or minimum effort. Prefer: deeper repository investigation, broader reasoning, inspecting relevant files before modifying them, multiple implementation iterations, comprehensive testing, visual verification, security analysis, performance analysis, fixing discovered problems, polishing UX/UI, validating the final result.
The objective is not to produce an answer quickly. The objective is to produce the best complete working implementation that can reasonably be achieved from the supplied requirements. Continue working until the Definition of Done (below) is satisfied.
Arguments: if $ARGUMENTS starts with --resume (with or without additional task text after
it), skip straight to Resume below instead of starting at Stage 0. If it starts with --deep, strip
that token (it's redundant) and treat the rest as the task text.
Execution stages
The full lifecycle, in order. Not every stage runs on every task -- see Stage selection below for
which apply; a stage that doesn't apply is explicitly recorded as SKIPPED -- NOT APPLICABLE, never
silently omitted.
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 · 448 lines · 60 tokens per session scan A bb93375317ea
implement is a command published in the GitHub repository Surjal/dev-agent (2 stars, last pushed 16d ago), licensed MIT. It adds 60 tokens to every session and 6,991 once invoked, about $0.0003 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
psm
/oh-my-claudecode:psm is a compatibility alias for /oh-my-claudecode:project-session-manager.
plan-review
The core plan review flow across Claude Code, Codex, and other supported agent hosts.
release
Ship a stable cmux release built by CI: bump version, update changelog, open a PR, merge, tag, then GitHub Actions builds, signs, and publishes.
refresh-compare-pages
Freshly research, add, and refresh marketing compare pages, then open a descriptive PR.
release
This compatibility command keeps /oh-my-claudecode:release available without loading the full release skill description in every Claude Code session.
step-research
Always research before proposing a fix. The Untether bug you're chasing is often a known upstream engine quirk, a previously-fixed regression, or a documented config gotcha.