agentleFS
l-gevityl-gevity/l-gevity-skills/.claude/skills/push-out/SKILL.md

push-out

Skill88 starsChanged 3 days agoMITView on GitHub

Moves recurring DevOps and operational work out of individual memory, manual execution, ticket queues, and team-local practice into documented procedures, repo standards, shared platforms, self-service controls, and adaptive feedback loops. Use when reducing toil, designing an improvement roadmap, deciding what to standardize or automate, assessing where operational work currently lives, pruning prose documentation duplicated by code/config/tests/policy-as-code/architecture-as-code, or asking how to move work from humans into durable systems.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 3 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do not push work to Rank 4 or 5 unless the lower-rank foundations exist or are created in the same change.

    line 78
  2. Distinctive:

    Question necessity: Delete work that no longer serves a real purpose.

    line 90
  3. Very specific:

    If many candidates compete, apply system-optimization to find the constraint before improving adjacent work.

    line 105
  4. Distinctive:

    Documentation pruning pattern: when prose documentation repeats behavior, rules, setup, policy, or architecture that is already enforced or …

    line 120
  5. Very specific:

    When the move concerns deployment safety, apply ci-cd-reliability-architecture.

    line 128
  6. Very specific:

    When the move concerns check placement, apply defect-shift-left.

    line 129

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

Push-Out

Purpose: Move recurring operational burden outward from fragile human execution into durable system capability.

Improvement Trio

  • defect-shift-left: move defect detection earlier.
  • push-out: move recurring operational work outward.
  • bring-down: move bespoke code down into reusable capability.

Core Directives

  1. Push work, not responsibility. The owner remains accountable; the execution path moves into a more durable layer.
  2. Evidence before rank. Assign no push-out rank without observed proof.
  3. Delete before pushing. Remove unnecessary work before documenting, standardizing, automating, or platforming it.
  4. Standardize before automating. Automation over ad hoc practice industrializes confusion.
  5. Self-service needs guardrails. A platform without the guardrail set — policy, validation, permissions, audit, observability, and rollback — exports toil to users.
  6. Feedback closes the loop. Dashboards and metrics are not improvement until they trigger action.
  7. Executable sources beat prose duplicates. When code, config, tests, schemas, CI, policy-as-code, or architecture-as-code already define the operational truth, keep prose only for intent, ownership, rationale, external constraints, trade-offs, rollback notes, and links to that source.

1. Push-Out Ladder

The ladder names where recurring operational work currently lives.

RankLocationWork lives inEvidence
0Individual memoryHeroics, tribal knowledge, manual clicks"Ask Alice", no runbook, no repeatable input/output
1Team procedureRunbook, checklist, documented handoffAnother team member can repeat it manually
2Repo standardTemplate, script, CI job, policy, config conventionThe repo enforces or strongly guides the path
3Shared platformGolden path, reusable workflow, managed internal primitiveMultiple repos/teams consume the same capability
4Self-service controlGuardrailed workflow developers can run without ops handoffUsers trigger it safely; the guardrail set exists (Directive 5)
5Adaptive systemMetrics and feedback change the systemThresholds, reviews, or automation drive continuous improvement
push-out distance = target rank - current rank

Higher is not always better. Choose the target by frequency, risk, toil cost, blast radius, and number of dependent teams.


2. Target-Rank Heuristics

ConditionMinimum target
Rare, low-risk internal taskRank 1
Repeated task inside one repo/teamRank 2
Repeated task across repos/teamsRank 3
Frequent developer-facing request or ticket queueRank 4
High-volume, high-risk, regulated, or incident-prone pathRank 5

Do not push work to Rank 4 or 5 unless the lower-rank foundations exist or are created in the same change.


3. Push-Out Protocol

  1. Define scope. Name the product, repo, platform, teams, environments, and time window.
  2. Inventory recurring work. Find manual steps, tickets, approvals, incident repeats, deploy chores, dashboard checks, and hand-maintained config.
  3. Question necessity. Delete work that no longer serves a real purpose.
  4. Locate current rank. Assign rank 0-5 with evidence.
  5. Choose target rank. Justify by frequency, risk, blast radius, compliance, toil cost, and dependency count.
  6. Compute distance. Target rank - current rank.
  7. Choose the next move. Emit one action that advances exactly one rank.
  8. Gate and retire. Prove the new path works, then remove same-scope manual duplicates.

Prioritize by:

priority = push-out distance x frequency x risk x blast radius x toil cost

If many candidates compete, apply system-optimization to find the constraint before improving adjacent work.


4. Move Patterns

MoveUse whenAction
0 to 1Knowledge is tribalWrite owner, inputs, outputs, runbook, rollback
1 to 2A runbook repeatsConvert to script, template, CI job, config schema, or policy
2 to 3Many repos copy the same practiceExtract shared workflow, platform primitive, or golden path
3 to 4Platform team is still a ticket queueAdd self-service UI/API/CLI with the full guardrail set (Directive 5)
4 to 5Self-service exists but does not improveAdd SLOs, trend review, alert thresholds, incident learning, and removal loop

Documentation pruning pattern: when prose documentation repeats behavior, rules, setup, policy, or architecture that is already enforced or derivable from code, config, tests, schemas, generated output, CI, policy-as-code, or architecture-as-code, do not create a second source of truth. Verify the executable source covers the same scope, then replace the prose duplicate with the smallest useful note: why it exists, who owns it, where the executable source lives, and which external constraint or trade-off is not visible from the code.

When the move concerns deployment safety, apply ci-cd-reliability-architecture. When the move concerns check placement, apply defect-shift-left. When the move concerns duplicated custom implementation, apply bring-down. When the question is which artifact owns the fact rather than whether prose duplicates an executable source, apply zero-copy-requirements.


5. Anti-Patterns

Anti-patternCorrection
Automating an undocumented processDocument and standardize first
Platform team as ticket queuePush to guardrailed self-service
Self-service without policy or rollbackAdd the full guardrail set (Directive 5)
Dashboard called improvementDefine threshold, review cadence, and action
Tool adoption treated as maturityScore the operational outcome, not the product installed
Golden path with no adoption signalMeasure usage, escape hatches, and support load
Manual approval called governanceReplace with policy-as-code where technically possible
Keeping manual duplicate foreverRetire same-scope duplicate after proof
Prose repeats an executable source of truthKeep intent/rationale/ownership/link; delete the repeated mechanics

6. Output Contract

Emit results in this shape:

Scope:          <product/repo/platform/team/environment/time window>
Mode:           Assessment | Improvement | Roadmap
Decision:       Keep manual | Delete | Document (→1) | Standardize (→2) | Platformize (→3) | Self-service (→4) | Adaptive (→5)
Summary:        <2-4 sentences: main toil source, best next push, key risk>
Verification:   <metrics, logs, workflow search, runbook check, or Not run + reason>

Work map:
| Work item | Current rank | Target rank | Distance | Evidence | Decision | Confidence | Next action |
| --------- | ------------ | ----------- | -------- | -------- | -------- | ---------- | ----------- |

Priorities:
| Rank | Work item | Why now | Push-out move | Validation | Duplicate to retire |
| ---- | --------- | ------- | ------------- | ---------- | ------------------- |

Gaps:
<Missing evidence, unknown ownership, absent metrics, excluded work, or risks>

7. See Also

  • defect-shift-left - move defect detection earlier.
  • bring-down - move bespoke code down into reusable capability.
  • system-optimization - bottleneck and waste analysis before improvement.
  • ci-cd-reliability-architecture - pipeline safety and deployment reliability patterns.
  • zero-copy-requirements - which artifact owns a fact, before pruning the duplicate.
  • continuous-improvement - promotion of recurring findings into skills, checks, or templates.

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.