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/doodledood/manifest-dev/definenpx skills add doodledood/manifest-dev --skill definegit clone --depth 1 https://github.com/doodledood/manifest-devWhat 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 | $0.00068 | $0.08098 |
| Opus 5 | $0.00034 | $0.04049 |
| Sonnet 5 | $0.00014 | $0.01620 |
| Haiku 4.5 | $0.00007 | $0.00810 |
Grade A, and why
define 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 — 220 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Where the manifest goes, and what runs first
Encode the conversation's shared understanding as a Manifest at ~/.manifest-dev/manifests/manifest-{ts}.md (create the dir; ~ = $HOME / %USERPROFILE%) — a durable home so manifests survive OS temp cleanup across multi-day work. Fall back to a writable temp path (/tmp/, else $TMPDIR / %TEMP%) only when the home directory isn't writable. If the transcript lacks shared understanding, invoke manifest-dev:figure-out first; propagate --autonomous when invoked from /auto or /do's amendment path. Pre-flight: if --babysit <pr-url>, load references/BABYSIT_MODE.md and follow its synthesis flow; if $ARGUMENTS contains a manifest file path, amend (see below); else fresh.
Encoding discipline
figure-out reaches shared understanding of the problem; /define handles manifest-specific encoding judgment calls — invariant vs process guidance, AC scope and pass threshold, gate kind. Surface the load-bearing encoding decisions briefly with a recommended answer before encoding; auto-decide the rest and mark (auto) + matching ASM. The manifest is the acceptance contract — what the user accepts as "I'd ship the outcome of executing this."
Problem first. Intent opens with the Problem — one specific story of what breaks or grates today, not an abstract need — because it is the baseline the rest of the manifest is written against: gates are pitched at what would actually relieve it, and whoever judges the result later weighs the work against that baseline rather than against an ideal. Appetite follows: the size of change the problem is worth, bounding complexity and surface rather than time or tokens, and the criterion for how many Deliverables to cut and how much to gate beyond what task files supply. Out of bounds records what the work deliberately leaves alone, and each bound is stated exactly once: one that must hold is written as the Global Invariant that binds it and appears nowhere else, while the rest stay a plain list that gains no shadow gate. A bound written twice is two texts that can drift apart, and only one of them binds. Elicit whatever the conversation left unset — these are encoding decisions, not re-investigation. A session that cannot name a pain has found a stop signal rather than an empty field: return to manifest-dev:figure-out, or conclude there is nothing worth building. With no user to ask, invoke manifest-dev:figure-out --autonomous to find the pain; only when that comes back empty, carry the thinnest Problem the evidence honestly supports with an ASM-* recording its thinness — and prefer returning no manifest to inventing one. All of that governs a fresh session. An amendment derives the fields from the manifest's own content plus the amendment's context and records a thin result as an ASM-*; it never halts, since a run may already be in flight and the manifest already exists.
What ships with it
28 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.
- references/BABYSIT_MODE.md 6.7 KB
- references/DOCS-STYLE-REFERENCE.md 15 KB
- references/MULTI_REPO.md 5.0 KB
- references/PROSE-FLOOR-REFERENCE.md 8.5 KB
- references/WRITING-REFERENCE.md 9.1 KB
- tasks/BLOG.md 708 B
- tasks/BUG.md 3.4 KB
- tasks/CODING.md 5.8 KB
- tasks/DOCUMENT.md 1.9 KB
- tasks/FEATURE.md 2.1 KB
- tasks/MARKETING.md 3.0 KB
- tasks/PR_LIFECYCLE.md 3.6 KB
- tasks/PROMPTING.md 4.2 KB
- tasks/PROSE_FLOOR.md 2.0 KB
- tasks/REFACTOR.md 1.0 KB
- tasks/references/research/agentic-systems.md 26 KB
- tasks/references/research/ai-failure-modes.md 27 KB
- tasks/references/research/epistemology.md 32 KB
- tasks/references/research/fact-checking.md 32 KB
- tasks/references/research/information-science.md 28 KB
- tasks/references/research/intelligence-analysis.md 14 KB
- tasks/references/research/README.md 32 KB
- tasks/references/research/systematic-reviews.md 24 KB
- tasks/research/RESEARCH.md 9.6 KB
- tasks/research/sources/SOURCE_WEB.md 4.0 KB
- tasks/TECH_DESIGN.md 2.0 KB
- tasks/UI.md 1.9 KB
- tasks/WRITING.md 1.4 KB
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 · 220 lines · 68 tokens per session scan A 341fdc884b9b
define is a skill published in the GitHub repository doodledood/manifest-dev (72 stars, last pushed 3d ago), licensed MIT. It adds 68 tokens to every session and 8,098 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 2026-08-30.
Other skills, from other repositories
external-context
Invoke parallel document-specialist agents for external web searches and documentation lookup.
security-ownership-map
Analyze git repositories to build a security ownership topology (people-to-file), compute bus factor and sensitive-code ownership, and export CSV/JSON for graph databases and visualization. Trigger only when the user explicitly wants a security-oriented ownership or bus-factor analysis grounded in git history (for…
moai-ref-ui-polish
UI polish and interface-completion reference: the small visual details — concentric border radius, optical alignment, shadow-vs-border, motion easing, typography smoothing, tabular numbers, icon stroke weight, hit areas — that separate polished interfaces from generic ones. Agent-extending skill that amplifies…
cross-task-learner
Enable agent loops to learn from similar past tasks and share patterns across loops.
ralph-specum-requirements
This skill should be used only when the user explicitly asks to use $ralph-specum-requirements, or explicitly asks Ralph Specum in Codex to run the requirements phase.
devkit-create-command
Create a new slash command with AI-guided behavior definition.