agentleFS
ravidsrkravidsrk/orca-fleet/skills/modernize-it/SKILL.md

modernize-it

Skill2 starsChanged 20 days agoMITView on GitHub

Bring a dependency / framework / platform surface current, safely — CI green at every merge. Inventory outdated + advisories with reachability triage → order by a compatibility graph → upgrade one dep or coherent group per PR, adapting call sites (code-level expand/migrate/contract) to get off old majors while keeping CI green → re-inventory until every major is current or pinned-with-a-reason. Use when "update the dependencies", "upgrade everything", "framework migration", "get off the old major", "move from Postgres 12 to 16", or an unattended dependency-currency run. Not for stateful DB schema/data migration across deploys (hand that to migrate-it), ordinary feature delivery (ship-it), or advisory exploit proof (harden-it).

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every outdated dep: upgraded+merged with CI green (the green run referenced) OR pinned-and-parked with a written reason + human ref.

    line 89

≈ “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

modernize-it — every major current or parked, CI green the whole way

You are the COORDINATOR. The unit is a COMPATIBILITY GRAPH node, not a finding: ordering, ecosystem grouping, lockfile contention, code-level adaptation, and rollback constraints dominate — PR-per-outdated-package is often actively WRONG. Composes remediate-finding, acceptance-review, risk-review (data-migration lens as a REVIEW SIGNAL, not an execution engine), runtime-prove, research-brief (INVENTORY reads the upstream changelog, not a summary of it), resolve-conflict (lockfile and hot-file chains), compound-learn; rides merge-serialization, reviewed-sha-freshness, dispatch-lifecycle, liveness-resume, evidence-manifest, ledger-contract, attention-budget, gate-classification (a pin decision needs a human ref), sandbox-policy (a bumped package's install hook is untrusted code — the run's single most dangerous execution surface). Worker TASK pack: addy — never co-mount a second router.

DEFERRED READS, never at activation: agent-brief.md at the FORCED-MIGRATION CHECK, when a handoff to migrate-it is written.

Scope boundary: this mission owns DEPENDENCY/FRAMEWORK CURRENCY (bump → adapt call sites → CI green). STATEFUL DB schema/data migration across deploys is a different unit, state machine, and proof — hand a brief to migrate-it, which owns each deploy-gated phase. Its plan places the dependent upgrade after expansion, preserves dual writes through backfill/switch, and contracts only after the upgraded code is deployed, stable and no longer uses the old shape. Track the handoff until its evidence returns; keep code-only upgrades here.

Two terminal outcomes

  • CURRENT — every dep on a current supported version, zero reachable unaddressed advisories.
  • CURRENT-WITH-PINNED (degraded) — all upgradable deps current, ≥1 pinned-and-parked (breaking upstream / dropped platform / unresolved conflict) with a human ref. Never reported as CURRENT.

"Current supported" authority (record it — registry-latest is NOT the truth)

The target is the version the PROJECT supports, by: an explicit project constraint (engines/peerDeps/ support policy), then the dependency's own published support/EOL policy (LTS/security window), then registry-latest only when neither exists. Record the chosen authority and its rationale in the unit evidence manifest. A dep on a still-supported older major is CURRENT.

Pipeline

INVENTORY (outdated + advisories; read the CHANGELOG not the version delta; reachability triage —
  dev-only/unreachable transitive is lower priority; never `audit fix --force`)
  → BOOTSTRAP integration BASE (runtime/scripts/preflight.py --mission modernize-it --base <BASE> --fork-point <sha
    recorded in the ledger header at BASE creation>; BASE ≠ default — dispatch-lifecycle.md)
  → order: security-critical-reachable → patch/minor (batch coherent groups) → majors (one per PR)
  → UPGRADE waves (one dep/coherent-group per PR: bump → adapt call sites, adding code-level
    deprecation shims / dual-run APIs where a major needs them → CI GREEN is the gate; lockfile
    regenerated not hand-edited; verify provenance on registry/maintainer change and install with
    lifecycle scripts disabled until provenance is verified — a bumped package's install hook is untrusted
    code (risk-review security lens). The lockfile is a hot mount-point file: upgrade PRs form a
    merge CHAIN (merge-serialization), regenerated at each rebase — and a rebase voids the review
    (reviewed-sha-freshness), so size coherent groups to keep the chain short)
  → FORCED-MIGRATION CHECK (risk-review data-migration lens): if an upgrade forces a stateful DB
    schema/data change, hand the migration and dependent upgrade to migrate-it as an `agent-brief`
    phase brief: expand → compatible upgrade/dual-write → backfill → switch → zero use → contract.
    Track the handoff in the inventory; migrate-it places the upgrade between expand and contract,
    with compatibility proven at each phase. Code-only upgrades continue through REVIEW below.
  → build-blind REVIEW (acceptance-review) → RUNTIME-PROVE (drive real entry points — green CI misses
    lazy imports and env-dependent init) → LAND
  → RE-INVENTORY → loop → outcome → REFLECT (`compound-learn`)

Convergence proof

Every outdated dep: upgraded+merged with CI green (the green run referenced) OR pinned-and-parked with a written reason + human ref. Every upgrade unit carries an EXECUTED negative control (evidence-manifest §1-§2): its criterion-bound command asserts the resolved version (lockfile/installed) is the target and runs the tests over the adapted call sites — for an advisory, its reachability probe — and reverting the unit turns it RED. A unit whose revert stays green has not shown the suite touches the upgrade. Every merge kept CI green (check the merge commits' checks — a red pipeline never landed). Every upgrade that FORCED a stateful DB migration has a migrate-it handoff tracking the dependent upgrade between expand and contract; unresolved handoffs remain parked in the inventory with reason and human ref. Completed handoffs return phase/deploy evidence before the node can count as current — never silently run inside this loop. Advisory scan re-run clean (or each remaining one parked with reachability rationale). Final inventory pasted. Manifest names CURRENT or CURRENT-WITH-PINNED.

Ledger + supervision

Header per liveness-resume.md: RUN · COORDINATOR · BASE · FORK_POINT · T0 · SOURCE · WIP (- if N/A; SOURCE = inventory digest). Rows include Orca task id + dep/group fields (target · PR · CI · pin/ handoff). Stalls → WATCH; death → RESUME scoped to header coordinator + ledger task ids, git-verified.

Anti-patterns

audit fix --force / mass-bump. Running a cross-deploy DB schema/data migration inside a currency loop (hand it to migrate-it). Rename-in-place code migrations (breaks dual-running deploys). Landing a red CI "to fix next PR". Bumping a major without its changelog. Dropping a compat shim without a gate. Hand-merging a lockfile conflict. Letting a just-bumped package's install hook run before its provenance is checked.

Related

migrate-it (owns stateful schema/data change across deploys), ship-it (ordinary feature delivery), clean-sweep, harden-it (advisory exploit proof), review-it (data-migration lens as a per-diff review).

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.