agentsclimarketplace

Ship

Skill SamyakJhaveri/loam/seed/.claude/skills/ship

Orchestrates the full shipping pipeline in strict order. Runs session-critique, then validate (both waves), then commit, then PR. Use when work is complete and ready to ship. Enforces ordering to prevent premature commits and skipped critiques. Accepts optional argument 'critique-only' to run just the critique step. NOT for mid-implementation checks (use /validate), code review without shipping (use /multi-review), or committing without the full pipeline (use /commit directly only if critique + validate already passed).From its SKILL.md

Install
npx -y skills add SamyakJhaveri/loam --skill ship

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.2 KB, 611 tokens by cl100k_base, as published. Nobody here has run it

Ship Pipeline

Strict-ordering orchestrator. Runs four stages in sequence, halting on failure.

Arguments

  • /ship (no args) → all 4 stages
  • /ship critique-only → stage 1 only (session-critique without proceeding to validate/commit/PR — useful for mid-session quality checks)

Hard Rules

  1. Never skip or reorder stages. The sequence is: critique → validate → commit → PR.
  2. If stage 1 or 2 fails, do NOT proceed to stage 3. Fix findings first.
  3. Each stage uses the existing skill's full logic. Do not reimplement any step inline — invoke the skill.
  4. Report status between stages. After each stage completes, state what passed and what's next.

Pipeline

Stage 1: Session Critique

Invoke /session-critique.

This spawns an advisor-pattern agent team that adversarially reviews all work in the current session against the decisions the user made during it. It surfaces findings for decision-drift, regressions, over-engineering, dangling references, and scope hygiene.

Gate: All BLOCK/HIGH/MEDIUM findings must be resolved (fixed or explicitly dismissed by user) before proceeding; an unresolved BLOCK halts the pipeline. If the user dismisses a finding, record the dismissal reason.

If critique-only was passed: Stop here. Report findings and exit. Do not proceed to Stage 2.

Stage 2: Validate

Invoke /validate (full — both waves, NOT quick).

This runs the two-wave validation loop: deterministic checks, rule-based tests. Must produce .validation_passed sentinel with waves_passed=2.

Gate: All waves must pass. On failure, enter the fix loop per .claude/rules/validation-loop.md. Max 3 iterations. After 3 fails, halt and escalate to user.

Stage 3: Commit

Invoke /commit.

Pre-check: Confirm .validation_passed exists and waves_passed=2 before proceeding. If sentinel is missing or stale, return to Stage 2.

Split commits by scope if the session produced multiple logical changes. Never bundle unrelated changes into a single commit.

Stage 4: PR

Invoke /pr.

Push to remote and open a GitHub PR with summary and test plan.

Post-check: Report the PR URL. Pipeline complete.

Failure Handling

FailureAction
Stage 1 finds HIGH/MEDIUM issuesFix findings, re-run Stage 1
Stage 2 validation failsEnter fix loop (max 3 iterations), re-run Stage 2
Stage 3 commit blocked by pre-commit gateReturn to Stage 2 (sentinel likely stale)
Stage 4 push failsReport error, do NOT force push
Any stage fails 3 timesHalt pipeline, escalate to user

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most review quality skills give in 611 tokens

Counted across 1,048 of the 1,783 authors here whose files we hold, read 2026-08-07

  • Ask questions one at a timein 81 of 1048, across 64 files
  • Provide a recommended answer for each questionin 73 of 1048, across 50 files
  • Explore the codebase instead of asking answerable questionsin 66 of 1048, across 42 files
  • Resolve dependencies between decisions one-by-onein 42 of 1048, across 17 files
  • Interview the user relentlessly about the planin 38 of 1048, across 13 files
  • Order findings by severityin 31 of 1048
  • Resolve each branch of the decision treein 27 of 1048, across 5 files
  • Run a grilling sessionin 26 of 1048, across 5 files
  • Update CONTEXT.md immediately when a term is resolvedin 26 of 1048, across 11 files
  • Propose precise canonical terms for vague languagein 25 of 1048, across 7 files
  • Create documentation files lazilyin 24 of 1048, across 5 files
  • Assign severity to every findingin 24 of 1048

Said here and by no other author read

  • run stages in order: critique then validate then commit then PR
  • halt on failure
  • invoke existing skills for each stage
  • do not reimplement any stage inline
  • report status between stages
  • resolve BLOCK HIGH or MEDIUM findings before proceeding

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 326,144. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.