agentleFS
scumunnascumunna/programmatic-skills/skills/bulk-change-governance-and-rollback/SKILL.md

bulk-change-governance-and-rollback

Skill1 starsChanged 5 months agoMITView on GitHub

Govern cross-platform bulk edits with scope control, before-state exports, diff review, canary rollout, maker-checker approval, verification, and rollback evidence. Use for SDF uploads, editor imports, API batches, spreadsheet uploads, campaign clones, budget updates, targeting edits, taxonomy fixes, and any large account change.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    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:

    If any gate fails, do not apply the full change.

    line 36
  2. Distinctive:

    Canary or validate: Use validate-only, preview, editor check, or one small slice first when the platform supports it.

    line 53
  3. Distinctive:

    Apply and read back: After apply, export or query the changed entities.

    line 56
  4. Very specific:

    Treat these as desk-lead or senior approval candidates through approval-and-escalation-governance:

    line 63
  5. Distinctive:

    Never bulk-edit active brand-safety controls without explicit approval.

    line 104
  6. Distinctive:

    If the result file has partial failures, the change is not done.

    line 105

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Bulk change governance and rollback

Bulk changes are where a good operator can save hours or break an account in one upload. This skill gives the cross-platform control pattern for bulk edits. It complements change-management-and-incident-response: that skill covers the broad incident and change runbook; this skill turns a proposed bulk edit into an auditable change package with rollback.

When to use this skill

  • A spreadsheet, editor import, SDF, API batch, platform bulk upload, campaign clone, or mass edit is proposed.
  • A change touches many campaigns, line items, ad groups, creatives, tags, budgets, bids, targeting settings, or taxonomy fields.
  • The user asks for rollback, four-eyes, dry run, canary, change approval, or "how do we do this safely".
  • A previous bulk change caused an incident and the team needs a safer pattern.

Bulk-change gates

GatePass condition
IntentThe business reason and requested field changes are written down
ScopeEntity IDs are enumerated and counted
SnapshotBefore-state export exists for every changed entity
DiffOnly intended fields changed
ApprovalMaker and checker are different people
CanarySmall slice applied or validate-only path passed where available
VerificationPost-change export matches expected state
RollbackExact rollback steps and artifacts are stored

If any gate fails, do not apply the full change.

Core process

  1. Write the change intent. State who requested it, why, which fields change, what result is expected, and what must not change. "Update bids" is not enough. "Increase max bid from $12 to $14 on these 18 CTV line items because deal floors increased" is actionable.
  2. Enumerate the scope. List exact entity IDs and names. Count them. Compare the count to the request. If the file contains more rows than expected, stop.
  3. Export the before-state. Pull the exact entities and fields before editing. Store it as the rollback artifact. No snapshot means no rollback.
  4. Build the proposed file or batch. Change only the intended fields. Preserve IDs, parent IDs, flights, budgets, status, targeting, tracking, and safety controls unless they are part of the approved change.
  5. Review the diff. Compare proposed state to before-state. The checker verifies changed fields, unchanged fields, row count, parent relationships, budgets, dates, statuses, and tracking references.
  6. Canary or validate. Use validate-only, preview, editor check, or one small slice first when the platform supports it. If no canary path exists, require desk-lead approval before full apply.
  7. Apply and read back. After apply, export or query the changed entities. Reconcile success count, warnings, rejected rows, and final state. Do not rely on "upload complete".
  8. Document rollback. Store before-state, proposed file, result file, approval, final export, and rollback instructions in the ticket.

High-risk changes

Treat these as desk-lead or senior approval candidates through approval-and-escalation-governance:

  • Budget shifts, daily caps, pacing mode, bid strategy, max bid, or tCPA or tROAS targets.
  • Activation status, pause, archive, delete, or flight date changes.
  • Brand-safety, suitability, verification, sensitive-category, or inventory-source changes.
  • Targeting expansions, geo changes, audience broadening, or exclusion removal.
  • Tracking, Floodlight, pixel, UTM, landing-page, click macro, or redirect changes.
  • Taxonomy changes that affect reporting joins, billing, or client-visible naming.

Rollback package

Bulk change package
Ticket:
Requester:
Maker:
Checker:
Approver:
Platform:
Advertiser:
Entity count:
Fields changed:
Business reason:
Before-state export:
Proposed file:
Diff summary:
Canary or validation result:
Full apply time:
Result export:
Verification:
Rollback file or steps:
Rollback owner:

Decision rules

  • The rollback must be executable by someone who did not make the change.
  • A clone is a bulk change. It can duplicate bad settings just as fast as an upload.
  • A field left blank by an upload may clear a platform value. Treat blanks as changes until the platform proves otherwise.
  • Never bulk-edit active brand-safety controls without explicit approval.
  • If the result file has partial failures, the change is not done. Reconcile the failed rows and decide whether to revert, retry, or hold.

Common pitfalls

  • Reviewing the proposed file without comparing it to a before-state export.
  • Making a manual UI tweak after the bulk file and leaving it out of the change record.
  • Applying a taxonomy cleanup that breaks dashboard joins or billing codes.
  • Trusting a platform import preview without reading back the final state.
  • Using "rollback" to mean "we could undo this later". Rollback needs a file or exact inverse operation now.

Related artifacts

  • templates/ops/bulk-change-plan.md
  • change-management-and-incident-response
  • approval-and-escalation-governance
  • cross-platform-taxonomy-governance
  • bulk-change-safety-gate loop

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.