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/ndisisnd/msg/intakenpx skills add ndisisnd/msg --skill intakegit clone --depth 1 https://github.com/ndisisnd/msgWhat 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 | $0.00239 | $0.02162 |
| Opus 5 | $0.00120 | $0.01081 |
| Sonnet 5 | $0.00048 | $0.00432 |
| Haiku 4.5 | $0.00024 | $0.00216 |
Grade A, and why
intake 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 — 101 lines — stays where its author put it; the contents beside it link to each section on GitHub.
intake
The front-door of the msg planning pipeline. Ideas and bugs enter here, get
interviewed into shape, and are recorded as graded rows in the root INTAKE.md
ledger — the living connective tissue between "things we want" and "PRDs that
exist." plan-pm reads those rows and drafts the PRD autonomously.
intake (capture + interview + grade → INTAKE.md rows) → plan-pm (drafts the PRD, stamps in-progress + prd mapping) → … → merge --production (stamps the row completed)
Usage
/intake [<idea text>]— capture one or more ideas/bugs intoINTAKE.md; pass the idea(s) as input, or be prompted. NL: "log an idea", "capture a bug", "add this to the backlog", "track this idea"./intake --update [<free text>]— list every un-shipped row (backlog+in-progress) in full, then edit the one you pick;#4 — goal should be …edits directly, but unclear input still gets a follow-up question. NL: "update that idea", "change the goal on #4", "fix the idea I logged"./intake --delete [#n]— list every row (includingcompleted), warn about what the removal breaks, then remove on explicit confirm. NL: "delete that idea", "remove row #4", "I logged that by mistake".
Mode dispatch. Resolve the mode once, at entry, from the arg string — --update or
--delete anywhere selects that mode and is stripped from the free text before target
resolution. The two are mutually exclusive; if both appear, say so and ask which. An ask
with no referent to an existing row is a capture, never an update. Then read exactly
one protocol and follow it end-to-end.
Hard refusals:
- Never drafts a PRD, never reads the codebase, never runs an analysis pass — in either mode. intake captures and grades;
plan-pmplans. A request to "plan this" or "write the PRD" hands off toplan-pm(recommend it; never invoke a full analysis here). - Never invents a fake-precise estimate (
~1,240 LOC,3.5 days). Grades are banded only (§ Grading). - Never writes
statusorprd, in either mode (../shared/refs/prd-lifecycle.md§ INTAKE.md row lifecycle). --updateeditsidea/goal/typeonbacklogrows only;gradeis never user-editable — it is re-derived on a material change, re-running capture's split gates. It refuses anin-progressrow (the PRD is the source of truth — use/plan-reviewor/plan-pm) and acompletedrow (historical record — capture a new idea instead); neither refusal ends the run.--updateand--deletenever scaffold a missingINTAKE.md— there is nothing to act on. Only capture mode scaffolds. A missingINTAKE-UPDATE.mdis never an error: absence means empty history; the file is lazy-created on the first log write.--updateis never destructive — removal is--delete's job, behind that mode's warning pass and confirm.--deletenever renumbers surviving rows (a gap is left soblocked-byreferences and log entries never silently repoint), and never deletes anything but ledger rows — not a PRD folder, file, or branch. It reports what it orphans; it does not clean up after itself.--deletenever proceeds without an explicit confirm, in any invocation form.
What ships with it
4 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.
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 · 101 lines · 239 tokens per session scan A 73186c7ed283
intake is a skill published in the GitHub repository ndisisnd/msg (6 stars, last pushed 15d ago), licensed MIT. It adds 239 tokens to every session and 2,162 once invoked, about $0.0012 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
skillspec-migrate
Complete a .agent.partial file by resolving TODO markers.
skill-writer
Review a SkillSpec .agent file for quality, correctness, and adherence to SkillSpec design principles.
skillspec-test
Execute test blocks from a SkillSpec .agent file.
skillspec-backport
Reconcile changes from a modified SKILL.md back into the .agent source file.
docs-planner
Identify documentation gaps and prioritize the docs backlog. Use when planning a docs improvement sprint, after signals surface repeated friction, when new SDK features ship without docs, or for periodic health assessment. Also triggers on "plan docs work", "what docs need writing", "prioritize the backlog", "docs…
goal-conductor
Own a long-horizon goal end to end - decompose it into work items, stand up one top-level session per item, patrol their state on a nudge loop, and decide each next round until the goal is met or a stop condition fires. Use when the user hands over a goal too large for one session ("clear the flaky-test backlog"…