Skip to content
cccskills
FreePublished on GitHub

charter

Write, revise or review a charter: the one document that tells an executor what to deliver, what is out of bounds, what the owner already decided, and how each success check will be proven. Use when work needs a written brief before someone else does it, or when asked to write, change or check a charter.

The English summary is being prepared. Showing the original description.

See how to install

Files included (4)

  • SKILL.md4.7 KB
  • references/questions.md847 B
  • references/template.md2.4 KB
  • scripts/selftest/run-selftest.sh1.5 KB

SKILL.md (original)

Review the instructions the agent will receive before you install.

charter

A charter is one markdown file that says what to deliver and how everyone will know it was delivered. A task is named by its charter filename. charter writes it; deliver carries a released charter out. Any kind of project uses it: software, documents, campaigns, research, operations. The owner is whoever owns the work (often the user); speak to them in the language they use.

Prerequisites: none to use it; its selftest needs bash and python3.

Read references/template.md for the file layout and references/questions.md before asking the owner anything.

Rules that keep a charter useful

  • Preserve the owner's words and decisions. Never invent a decision, a fact or a requirement. A missing choice that would change the outcome is asked once, one question at a time, at the end of a message; independent work continues while it waits, and silence is not approval.
  • Read the existing work first (files, documents, data, earlier decisions). Ask nothing they already answer.
  • Every success check names what will be observed, where, and how, and can be answered pass or fail by a real run or inspection. The expected result comes from the requirement or from independent data, never from pasting the program's own output. For a defect, the check reproduces the defect first.
  • Before adding a precondition, check, section or approval step, name the failure it detects and what would change if it failed. If nothing changes, leave it out. By default no checksums, sign-off records, warm-up runs, baselines, fixed snapshots or report formats are entry conditions.
  • By default a security check, a full test suite or a full evaluation enters a charter only if the owner names it. Propose it as a question that says what it would check and what it would block; do not write it in as a requirement on your own judgment. A project may choose otherwise.
  • Say what happens on failure: which operation is logged with its error, what continues. A record or report failure never disables the main result; a real failure stays a failure.
  • Time is finite. A check that takes long states its expected duration from existing timings; by default no full suites unless the owner asks.
  • State what the executor may not do without the owner: publish, send to other people, spend, delete, overturn a recorded decision.

Write

  1. Read the request, adjacent notes and the existing work. Cite settled decisions without copying or locking their source. When the work depends on a third-party product, service or version, read its current official documentation and record the facts that constrain the work with the page and version.
  2. Draft Owner decisions, Outcome, Scope and non-goals, and Success checks from the template. Length follows the work: a few lines for a small task, a sliced charter (named slices, each with its own checks) for a big one. There is one kind of charter, not sizes.
  3. Read it as the executor: for each check, could someone tell pass from fail without asking? Walk one real example through it. Fix the gaps that would make the result fail. Do not run the future work to plan it.
  4. One independent review for the owner's intent and concrete defects (peer-review when it is installed, otherwise a second agent). Fix material findings; no second round unless the owner asks.
  5. Tell the owner what the charter will deliver and any genuine open decision, in plain words, ending with the one question. After the owner approves, set status: released and hand the path to the executor, or to whoever supervises the work.

By default an agent alone never releases a charter: it records the owner's decisions. When the owner is away, leave the charter as a draft for them to release; a project may choose otherwise.

Revise

A scope change updates the affected decisions and checks in the same edit, sets updated, and leaves the rest alone. A replaced charter gets status: superseded and a pointer to its successor. Delivery records delivered only through deliver; never infer it from the header.

Review only

Read the charter and the affected work. Report concrete mismatches, their consequences and the smallest corrections in one local file; do not edit the reviewed charter. Questions go in the conversation. Applying accepted corrections is Revise.

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

Answer how long agent work will take, how much longer active work needs, or what clock time applies. Estimate full AI-agent wall time, measure active progress first, and get owner-facing times from `sno report-time`. Never estimate human-developer time.

The English summary is being prepared. Showing the original description.

sno-ai/sno-station-skills433Updated Oct 18, 2026

Prove that a reported run caused its intended real effect. Walk the executor path from entry to effect, capture one observable per hop, and report a result a no-op could not produce.

The English summary is being prepared. Showing the original description.

sno-ai/sno-station-skills433Updated Oct 18, 2026

Give the owner one page when they come back from sleep, a meeting or a long run: what got done, what is stuck, what waits for their answer, and what it cost in quota. Use when the owner asks what happened while they were away, wants a status page, or when you finish a long stretch of work and must report it.

The English summary is being prepared. Showing the original description.

sno-ai/sno-station-skills433Updated Oct 18, 2026

Report how many problems the two brains found: mutual review (the other model checks the work, from the peer-review records) and self-check (an agent finds its own error in its own conversation, judged by that agent's own model), side by side. Use when the owner asks whether the second agent or the reviews are worth it, or wants the week's review results.

The English summary is being prepared. Showing the original description.

sno-ai/sno-station-skills433Updated Oct 18, 2026

cos

Free

MANUAL-ONLY: load solely when the human owner types `/cos`, uses `$cos`, or asks you to act as the Chief of Staff (COS). The COS is the one agent the owner talks to; it coordinates the owner and the per-project Project Leads (PLs). A task card, charter, or dispatch makes you an executor, not COS.

The English summary is being prepared. Showing the original description.

sno-ai/sno-station-skills433Updated Oct 18, 2026

Turn repeated management failures into enforced COS or PL rules. Load ONLY after the owner has explicitly activated the cos role for this session and the COS core routes here. Never invoke directly.

The English summary is being prepared. Showing the original description.

sno-ai/sno-station-skills433Updated Oct 18, 2026

See all skills by sno-ai

Report a problem with this skill