agentleFS
wxxb789wxxb789/hoa/skills/research-stream-consolidation/SKILL.md

research-stream-consolidation

Skill3 starsChanged 53 days agoMITView on GitHub

Merge parallel research streams that cover the same topic into one project: inventory both trees, relocate a stream as a single unit, dedup by content identity exactly once, and repair the path-bound records the move invalidated. Use when two agents or models produced overlapping research and the deliverable must become one tree without losing provenance. Not for copying files between projects or for reconciling divergent code branches.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is 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.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use when two or more independent research runs cover the same topic (typically the same owner commissioned them from different agents or …

    line 20
  2. Very specific:

    If the evolved deliverable has a length/format contract, assemble and QA it with the discipline of contract-bound-deliverables.

    line 26
  3. Distinctive:

    Re-read moved files before editing them: After a path move the edit tool requires a fresh read at the new path.

    line 55

≈ “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

Research stream consolidation

Use when two or more independent research runs cover the same topic (typically the same owner commissioned them from different agents or models) and they must become one project: merge A into B, delete A, dedup B, then review the union and evolve the deliverable. This is a relocation + reconciliation problem, not a copy-paste: research repos bind artifact paths in manifests, receipts, study-control records and navigation docs.

If the evolved deliverable has a length/format contract, assemble and QA it with the discipline of contract-bound-deliverables; this skill covers the merge and reconciliation around it.

Procedure

  1. Inventory before moving anything. List both trees (find, du), read the target project's README/plan for its rules (what may be deleted, what must be kept), and grep the whole repo for the source path — study-control records, manifests, READMEs and wiki pages bind it. Every hit is a record to update or annotate afterwards.
  2. Relocate the stream as one unit. git mv <source-dir> <target-project>/<stream-name>/ preserves internal structure, relative paths and receipts; scattering files into the target destroys them. Run git mv and any git rm as separate commands: a compound git mv … && git rm … exits 1 because rename-staged files need -f for removal — check git status, then re-run the removal alone.
  3. Dedup by content identity — once. Hash the candidate duplicates in a single pass (sha256sum), keep one representative, git rm -f the rest. One identity pass decides; re-hashing to "confirm" the same files later is stalling. Respect the project's keep rules: files a plan says to retain (e.g. empty result/error outputs as run evidence) are not dedup targets even when byte-identical.
  4. Record the relocation on both sides. Artifact side: what moved, what was deleted, how to recover from git history. Study/control side: a note where the old path was bound, so path-bound records stay traceable. Then update navigation docs (README/MATERIALS) and re-resolve their links mechanically (parse every link, not just spot-check). Then re-run the study's validator and repair path/hash binding drift (repair the path and hash bindings in the study's control records) — a relocation note documents the move; it does not fix the bound records.
  5. Re-read moved files before editing them. After a path move the edit tool requires a fresh read at the new path; overwriting from pre-move context fails.
  6. Reconciliation review. Compare the two streams' item sets — they often barely overlap, and that is the merge value. Build a conflict table for claims that differ (same program, different status wording) and write the disposition: which stream's wording wins, and why. Figures the streams independently rejected stay rejected — independent convergence on a drop is corroboration, not a gap to fill.
  7. Evolve the deliverable from the union. Preserve each claim's original evidence level — appearing in two streams does not promote it (a summary-level item does not become claim-level by duplication). Put the provenance map (new reference id → source stream/claim), conflict dispositions and a not-included log in a sidecar memo, not in the deliverable. Re-run the deliverable's own QA script against the union corpus.

Pitfalls

  • Re-verifying after every step (re-hashing, re-checking). Establish integrity once, where it decides something (provenance, duplicate detection), then advance — the owner calls repeated hash loops out as stalling.
  • Resurrecting a figure both streams deliberately dropped.
  • Moving files without the study/control-side note — and stopping there: manifests and closeout records still point at the old path, and the study validator fails on the broken bindings; repair the chain (repair the path and hash bindings in the study's control records), do not just annotate.
  • Editing a moved file from a stale (pre-move) read.
  • Treating byte-identical files as mergeable content: keep one, delete the rest, and say so in the relocation record.

Supporting files

  • references/reconciliation-formats.md — conflict-disposition table, provenance map and not-included log formats for the sidecar memo.

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.