raci — declared by cat-harness — source held

Applies when. Who is involved in an activity, and how. Use it when a process or a breakdown exists and the question is participation — who answers for this, who must be asked first, who is told afterwards. Not a decision method: a one-off choice among options is kepner-tregoe, a recurring rule is dmn, certainty of evidence is grade, the record of a decision is madr, and situation analysis before any of them is swot. RACI answers who, never what or whether.

Origin. NO PRIMARY SOURCE IS HELD HERE, and the ingested one is SECONDARY. RACI is a responsibility assignment matrix from project-management practice. The text this node cites renders the four roles but does not originate them — it attributes them onward to PMI’s PMBOK Guide (2021), which is a paid standard nobody here has opened. The earlier Linear Responsibility Chart literature is the other commonly named ancestor. Both stay in §”The primary is still not held” as CANDIDATES TO FETCH, unverified, rather than asserted here as provenance. Rendering from an open secondary was the owner’s ruling of 2026-09-23 — the route swot took. See literature-search.

Ingested sources:

library/dusengumuremyi-2026-ai-mediated-raci · view · item page

Applied by skills:

raci

Node file: raci.md

RACI — involvement, in four kinds

Adopted on the owner’s ruling, 2026-09-22 (bean 2xfl): RACI is a methodology, not a house practice, so it has a node here. Until that ruling it existed only as a skill, and the methodology graph reported five nodes where a reader might reasonably expect it to report this one too.

⚠ The cited source backs ONE sentence of this node, and nothing else
evidence points at Dusengumuremyi (2026), ingested 2026-09-23. It earns its place only as a secondary rendering of the four roles — its §2.1, verbatim from sections/page-002.md:
a classic project management instrument delineating four roles: Responsible (doer), Accountable (owner), Consulted (input provider), and Informed (notified).
Everything else in that paper is rejected, and the rejection is load-bearing rather than fastidious. Its own Data Availability statement reads “No new datasets were generated or analyzed during this study” while its abstract claims 1,247 decision events and its §4.1 reports t(1245)=47.3, p<.001; its abstract’s headline 92.4 % detection rate appears nowhere in its findings, which say 78.4 %; and its running head claims Journal of International Business Studies “Vol. 1, 2026” while its own reference list cites that journal at vol. 53 in 2022. Bean thb1 holds the full assessment with every item checkable against the ingested pages.
So: cite this source for the four letters. Never cite it for a claim about what RACI achieves. No efficacy, latency, error-rate or risk-detection statement anywhere in this repository may rest on it.
That split is why evidence here is not the same kind of backing swot has. swot rests on two sources that survive reading whole; this rests on one paragraph of one paper that does not. check:methodology-evidence counts nodes with a resolvable source and will now count this one — the axis measures whether a citation resolves, not whether a reader would accept it, and conflating those is what this box exists to prevent.

The four letters

letter	means	direction
Responsible	does the work	may be several
Accountable	carries the decision; answers for the outcome	exactly one
Consulted	asked for input before	two-way
Informed	told after	one-way

The axis that makes it a method rather than a list is time and direction: C happens before and is a conversation; I happens after and is a broadcast. Collapsing them turns consultation into notification, which is the failure the letters exist to keep apart.

Exactly one Accountable

The method’s central constraint, and the one most often broken in practice. A chart with two A’s does not record a shared decision — it records that nobody established who answers for the outcome, and it reads as complete while doing so.

Zero is also a breach, once a row claims anything at all. A half-filled row is worse than an empty one for the same reason: the chart looks answered.

A row where the same party is both Accountable and Consulted is not a row with two facts in it. Asking yourself is not consultation, and that pairing is how consulted becomes a formality while the chart still reads complete.

What RACI is not

Not a decision method. A is who answers for a decision, not how it is reached. Reaching it is kepner-tregoe or dmn; recording it is madr.
Not an approval gate. Answering for an outcome and signing off on one are different acts, and a method that conflated them would let a chart stand in for a gate.
Not a permission model. What a party may do in general is a different question from who is involved in this activity, and the two cross-cut.
Not a party registry. RACI names roles in rows; it does not establish who exists. Something else has to say that first.

The primary is still not held

A secondary rendering is cited above; the primary is not, and the gap has narrowed rather than closed. literature-search’s outcome 3, recorded so the next agent does not repeat the search: nothing was found because nothing could be fetched. Measured 2026-09-22 — doi.org, pmi.org and every open-access aggregator fail through this container’s proxy; only raw.githubusercontent.com answers.

Candidates worth fetching, unverified and listed as leads rather than as provenance:

candidate	why it is a candidate	access
PMI, PMBOK Guide — the responsibility assignment matrix (RAM)	the most commonly cited formal treatment	not open access
The Linear Responsibility Chart literature (Cleland & King and successors)	commonly named as the earlier form RACI descends from	books, not open access

RACI and kepner-tregoe were in the same bucket, and RACI has left it. For both, the commonly cited sources are books and paid standards, so no open-access primary exists to fetch, and closing the gap needed either a copy or a decision to render from an open secondary — the route swot took. The owner made that decision for RACI on 2026-09-23. kepner-tregoe is still in the bucket, and the same two exits are still open to it.

How this platform applies it

As an overlay on declarations that already exist, never as a new table of names — and R is read rather than declared, because a BPMN lane already says who performs an activity.

That application, its three declared edge kinds, the cat-harness.processes:raci extension element, the check:raci gate, and the procedure for running a RACI exercise when initiating a project are in the skill: skills/process/raci/raci.md. The method lives here once; the skill names it and does not restate it.