A preserved method from https://github.com/mattpocock/skills at 3cca18b368ae, path skills/engineering/research, MIT. 22 of 15 source lines differ (147%), every difference claimed by an entry of the ledger with its reason. Entries: baseline-copies-2026-09-11, pull-2026-09-11, fold-2026-09-11, fold-walk-2026-09-11, vocabulary-owner-2026-09-12, cold-review-wording-2026-09-12.
greenline renders its own frontmatter: quoted name and description, the description from the manifest override where one existed, no upstream activation flag
name: research
name: "research"
description: Investigate a question against high-trust primary sources and capture the findings as a Markdown file in the repo. Use when the user wants a topic researched, docs or API facts gathered, or reading legwork delegated to a background agent.
description: "Investigate a question against high-trust primary sources and capture the findings as a Markdown file in the repo. Use when the user wants a topic researched, docs or API facts gathered, or reading legwork delegated to a background agent."
---
---
scopechangedcold-review-wording-2026-09-12
The skill's scope opens the body in its own voice: a stage (wayfinder, grill-with-docs or to-spec) dispatches it for a decision or task waiting on a factual answer, it uses that existing owner rather than recency, it supplies evidence without settling the decision, and a read-only request writes nothing; the diff pairs this paragraph with upstream's background-agent line, which continues, extended, in the next hunk (carried from fold-2026-09-11; the word workflow became stage, the internal refactor's step 1, ruling 2, after the cold review of the cutover, Prose 1, 2026-09-12)
Spin up a **background agent** to do the research, so you keep working while it reads.
A stage dispatches this method when a decision or task is waiting on a factual answer: wayfinder, grill-with-docs or to-spec names the question, and the answer goes back to that stage. Identify the decision or task waiting on the answer and use its existing owner; recency alone does not choose an initiative. The decision itself stays with the requesting stage: this method supplies cited evidence and does not settle the question. A read-only request returns its cited answer in the reply and writes nothing to the repository.
lifecyclechangedfold-2026-09-11
The background agent's reading is its own consultation under greenline's ledger: the requester cites the file it hands back and does not record the agent's sources as ones it read itself (the consumer block's rule that another agent's receipt is not its reading).
Spin up a **background agent** to do the research, so you keep working while it reads. Its reading is its own: cite the file it hands back, and do not record its sources as ones you consulted.
Its job:
Its job:
1. Investigate the question against **primary sources** (official docs, source code, specs, first-party APIs), not a secondary write-up of them. Follow every claim back to the source that owns it.
1. Investigate the question against **primary sources** (official docs, source code, specs, first-party APIs), not a secondary write-up of them. Follow every claim back to the source that owns it.
lifecyclechangedvocabulary-owner-2026-09-12
the handoff section: dispatched by wayfinder or to-spec, or by the request owner directly; grill-with-docs does not dispatch research (carried from fold-walk-2026-09-11; the word operator became owner where the copy names the workspace's person, and request owner became request holder (the internal refactor's step 1, ruling 10, 2026-09-12))
2. Write the findings to a single Markdown file, citing each claim's source.
2. Write the findings to a single Markdown file, citing each claim's source: the question, the supported findings, the counterevidence, and the limits of the sources.
3. Save it where the repo already keeps such notes; match the existing convention, and if there is none, put it somewhere sensible and say where.
3. Save it as the owning initiative's `research/RSRCH-NNN.md`, with the supporting-artifact frontmatter `.greenline/WORK.md` defines, or under `.greenline/work/evidence/TKT-NNN/research/` when a compact ticket owns the question, and say where.
## Handoff
Consumes: the decision or task waiting on the answer, from wayfinder or to-spec, or from the request holder
Produces: research/RSRCH-NNN.md in the owning initiative, or a file under .greenline/work/evidence/TKT-NNN/research/ for a compact ticket; nothing for a read-only request
Evidence at: that file, cited claim by claim
Returns to: the requesting stage, with the citation and the pinned revision
agents/openai.yaml
harnessremoved filebaseline-copies-2026-09-11
the renderer generates agents/openai.yaml from the manifest; the vendored copy is not projected
interface:
display_name: "Research"
short_description: "Research from high-trust sources"
The timeline
Each entry that touched this method, with the differences it claimed as they stood at its commit, read from the repository's history.
2026-09-11baseline-copies-2026-09-11
the cutover to an edited copy (ADR 0039): the composed output written as the copy, every difference from upstream claimed with the reason of the overlay that produced it
SKILL.md
harnesschangedbaseline-copies-2026-09-11
greenline renders its own frontmatter: quoted name and description, the description from the manifest override where one existed, no upstream activation flag
name: research
name: "research"
description: Investigate a question against high-trust primary sources and capture the findings as a Markdown file in the repo. Use when the user wants a topic researched, docs or API facts gathered, or reading legwork delegated to a background agent.
description: "Investigate a question against high-trust primary sources and capture the findings as a Markdown file in the repo. Use when the user wants a topic researched, docs or API facts gathered, or reading legwork delegated to a background agent."
lifecyclechangedbaseline-copies-2026-09-11
greenline prelude, to fold: identify the decision or task waiting on the answer and use its existing owner; a read-only request returns the cited answer without writes
**greenline prelude: evidence for an owned question.** Identify the decision or task waiting on this answer. Use its existing owner; recency alone does not choose an initiative. The shared work contract supplies a research artifact for initiative research or an evidence home for a compact ticket's supporting investigation. A read-only research request returns its cited answer without repository writes. The method's answer returns to the requesting agent within the same grant.
lifecyclechangedbaseline-copies-2026-09-11
greenline completion, to fold: retain question, findings, counterevidence, limits and citations; pin any revision a downstream decision consumes
## greenline completion: cited research
Retain the question, supported findings, counterevidence, source limitations, and citations where the work owns them. An initiative research artifact follows `.greenline/WORK.md`; a compact ticket can link its research evidence without acquiring an artificial initiative. Pin any artifact revision a downstream decision actually consumes. The contributor's ledger records the sources it really read; a summary received from another agent is not that agent's source consultation. Return the result to the original workflow.
agents/openai.yaml
harnessremoved filebaseline-copies-2026-09-11
the renderer generates agents/openai.yaml from the manifest; the vendored copy is not projected
interface:
display_name: "Research"
short_description: "Research from high-trust sources"
2026-09-11pull-2026-09-11
pin advance to 3cca18b: upstream moved with no change under the vendored paths
2026-09-11fold-2026-09-11
The prelude and completion are folded into the body: the scope opens the method, the findings file has one greenline home named at the step that saves it, and a Handoff section returns the citation to the requesting stage.
SKILL.md
scopechangedfold-2026-09-11
The skill's scope opens the body in its own voice: a workflow stage (wayfinder, grill-with-docs or to-spec) dispatches it for a decision or task waiting on a factual answer, it uses that existing owner rather than recency, it supplies evidence without settling the decision, and a read-only request writes nothing; the diff pairs this paragraph with upstream's background-agent line, which continues, extended, in the next hunk.
Spin up a **background agent** to do the research, so you keep working while it reads.
A workflow stage dispatches this method when a decision or task is waiting on a factual answer: wayfinder, grill-with-docs or to-spec names the question, and the answer goes back to that stage. Identify the decision or task waiting on the answer and use its existing owner; recency alone does not choose an initiative. The decision itself stays with the requesting stage: this method supplies cited evidence and does not settle the question. A read-only request returns its cited answer in the reply and writes nothing to the repository.
lifecyclechangedfold-2026-09-11
The background agent's reading is its own consultation under greenline's ledger: the requester cites the file it hands back and does not record the agent's sources as ones it read itself (the consumer block's rule that another agent's receipt is not its reading).
Spin up a **background agent** to do the research, so you keep working while it reads. Its reading is its own: cite the file it hands back, and do not record its sources as ones you consulted.
2026-09-11fold-walk-2026-09-11
the coherence walk: the dispatchers named are the ones that dispatch
SKILL.md
lifecyclechangedfold-walk-2026-09-11
the handoff section: dispatched by wayfinder or to-spec, or by the request owner directly; grill-with-docs does not dispatch research
2. Write the findings to a single Markdown file, citing each claim's source.
2. Write the findings to a single Markdown file, citing each claim's source: the question, the supported findings, the counterevidence, and the limits of the sources.
3. Save it where the repo already keeps such notes; match the existing convention, and if there is none, put it somewhere sensible and say where.
3. Save it as the owning initiative's `research/RSRCH-NNN.md`, with the supporting-artifact frontmatter `.greenline/WORK.md` defines, or under `.greenline/work/evidence/TKT-NNN/research/` when a compact ticket owns the question, and say where.
## Handoff
Consumes: the decision or task waiting on the answer, from wayfinder or to-spec, or from the request owner
Produces: research/RSRCH-NNN.md in the owning initiative, or a file under .greenline/work/evidence/TKT-NNN/research/ for a compact ticket; nothing for a read-only request
Evidence at: that file, cited claim by claim
Returns to: the requesting stage, with the citation and the pinned revision
2026-09-12vocabulary-owner-2026-09-12
The workspace's person is the owner: the word operator became owner where the copy names the workspace's person, and request owner became request holder (the internal refactor's step 1, ruling 10, 2026-09-12); every earlier claim on a re-measured hunk is carried forward under its own kind, with its authority where it had one.
SKILL.md
retiredchangedvocabulary-owner-2026-09-12
re-measured after the word change; its claim (lifecycle from fold-walk-2026-09-11) carries to the hunk that replaced it
2. Write the findings to a single Markdown file, citing each claim's source.
2. Write the findings to a single Markdown file, citing each claim's source: the question, the supported findings, the counterevidence, and the limits of the sources.
3. Save it where the repo already keeps such notes; match the existing convention, and if there is none, put it somewhere sensible and say where.
3. Save it as the owning initiative's `research/RSRCH-NNN.md`, with the supporting-artifact frontmatter `.greenline/WORK.md` defines, or under `.greenline/work/evidence/TKT-NNN/research/` when a compact ticket owns the question, and say where.
## Handoff
Consumes: the decision or task waiting on the answer, from wayfinder or to-spec, or from the request owner
Produces: research/RSRCH-NNN.md in the owning initiative, or a file under .greenline/work/evidence/TKT-NNN/research/ for a compact ticket; nothing for a read-only request
Evidence at: that file, cited claim by claim
Returns to: the requesting stage, with the citation and the pinned revision
2026-09-12cold-review-wording-2026-09-12
The word workflow is retired for stage in the product's speech (the internal refactor's step 1, ruling 2), and the cold review of the cutover (Prose 1, 2026-09-12) found it in the copy's opening paragraph; the paragraph now says a stage dispatches the method. The earlier claim on the re-measured hunk is carried forward under its own kind and reason.
SKILL.md
scopechangedcold-review-wording-2026-09-12
The skill's scope opens the body in its own voice: a stage (wayfinder, grill-with-docs or to-spec) dispatches it for a decision or task waiting on a factual answer, it uses that existing owner rather than recency, it supplies evidence without settling the decision, and a read-only request writes nothing; the diff pairs this paragraph with upstream's background-agent line, which continues, extended, in the next hunk (carried from fold-2026-09-11; the word workflow became stage, the internal refactor's step 1, ruling 2, after the cold review of the cutover, Prose 1, 2026-09-12)
Spin up a **background agent** to do the research, so you keep working while it reads.
A stage dispatches this method when a decision or task is waiting on a factual answer: wayfinder, grill-with-docs or to-spec names the question, and the answer goes back to that stage. Identify the decision or task waiting on the answer and use its existing owner; recency alone does not choose an initiative. The decision itself stays with the requesting stage: this method supplies cited evidence and does not settle the question. A read-only request returns its cited answer in the reply and writes nothing to the repository.