agentleFS
sufficiently-advanced-aisufficiently-advanced-ai/imi/skills/core/constitution-review/SKILL.md

constitution-review

Skill8 starsChanged 42 days agoMITView on GitHub

Produce a constitution review from the user's imi knowledge base — the standing rules, protocols, scope boundaries and commitments currently in force, with what is deferred, superseded or in conflict. Use when the user asks "what are our standing rules", "show me the constitution", "what have we decided as policy", "what's still in force for [client or project]", "what would I tell someone new about how we operate", or wants to audit drift between what was decided and what is happening. Works with any imi domain.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    All tools below are on the imi MCP server (possibly prefixed, e.g.

    line 19
  2. Distinctive:

    If no imi tools are available, say so and stop.

    line 20
  3. Distinctive:

    When in doubt, include it: hiding an active rule is worse than showing a closed one.

    line 49
  4. Very specific:

    Never present an unreviewed decision as settled policy.

    line 57

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Constitution Review

A constitution is not every decision ever made. It is the decisions that still govern what happens next: protocols, scope boundaries, commitments, policies. "We closed the SSO project" is a fact for meeting notes. "Change requests need written sponsor approval" is constitutional.

All tools below are on the imi MCP server (possibly prefixed, e.g. mcp__imi__get_constitution). If no imi tools are available, say so and stop.

1. Scope

  • Global — no organization or project named: everything.
  • Scoped — one named: resolve it with get_entity_by_name (or search_knowledge_graph) and keep its entity id. Scoped reviews also include global rules that apply to it.

2. Load the decisions

  1. get_constitution — the server's current decision constitution as Markdown: active, conflicting, stale and superseded decisions with owners, dates and governance authority (instruction-grade, evidence-grade, blocked). This is the backbone.
  2. list_decisions with state "active", then again with "stale" and "superseded" — the lifecycle states are computed server-side; trust them over your own date arithmetic. Add client_id when scoped and the domain has client entities; otherwise filter by reading.
  3. search_signals with signal_type "insight" (and the entity_id when scoped) — open flags that represent decisions not yet made.
  4. For any decision whose lineage matters (it replaced or was replaced by another), call get_decision with its id to read the supersession chain.

3. Classify

For each decision ask: does it govern future behaviour?

  • Standing rule — protocols, scope boundaries ("X is out of scope for Y"), ongoing commitments, policies, deferred decisions with a revisit date.
  • Situational fact — something that happened and is finished. Leave it out.

When in doubt, include it: hiding an active rule is worse than showing a closed one.

4. Respect authority (ADR-002)

imi separates instruction-grade decisions (a human confirmed them) from evidence-grade ones (extracted, not yet reviewed). Keep that line visible:

  • Mark each rule confirmed or unreviewed, from the authority the constitution reports.
  • Never present an unreviewed decision as settled policy. If the user wants one promoted, point them to the review queue or update_signal with a review action — do not confirm decisions yourself unless the user explicitly tells you to, item by item.

5. Write it

## <Organization / client / project> — Constitution
As of <today> · <N> decisions reviewed · <M> confirmed

### Standing rules
- <rule, present tense> — *confirmed|unreviewed* *(decided <date>, <source>)*

### Active commitments
Promises still in motion that bind the organization (not individual to-dos).

### Deferred
Decided but on hold, with the revisit date if known.

### Open flags
Insights that point at a rule nobody has made yet.

### Conflicts and drift
Rules that pull against each other, or later events that contradict a rule. State them
neutrally — they are things to resolve, not accusations.

Group standing rules under headings if there are many (engagement protocols, scope boundaries, commercial terms). Omit empty sections; never write "none". Dates are when the decision was made, not when imi recorded it.

Tone

A handbook, not a report: present tense, one sentence per rule. If it is getting long, situational facts slipped through — filter harder.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.