From gcp-hcp
Use when an implementation plan in gcp-hcp may have drifted from actual code, or when code may have drifted from the plan's design intent. Typically after a feature has landed or a PR has merged. Checks API surface, behavior, interfaces, and structure across repositories.
How this skill is triggered — by the user, by Claude, or both
Slash command
/gcp-hcp:fix-implementation-plan-driftsThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Implementation plans are living documents. When code ships, the plan should reflect what was built. This skill detects the gap and determines whether to update the plan or flag the code for review.
Implementation plans are living documents. When code ships, the plan should reflect what was built. This skill detects the gap and determines whether to update the plan or flag the code for review.
gcp-hcp-architecture skill to identify which repo owns each component. If the plan references repos beyond the user's selection, list them and ask if the user wants to include them.Drift is the gap between what the plan says and what the code does. It goes both ways:
Focus on meaningful differences — API contracts, behavior, architecture. Don't nitpick formatting or stylistic choices.
For each plan, verify these against actual code:
API surface — types, structs, or fields don't match actual definitions. Grep for every type and field mentioned. Compare names, tags, and markers. (e.g., plan references GCPNetworkSpec but code renamed it to GCPNetworkConfig)
Behavior — controller reconciliation logic, ownership patterns, or data flow differ from actual code. Read the actual reconcilers. (e.g., plan describes credentials managed in HO layer but code manages them in CPO)
Interfaces — CLI commands, flags, or workflows don't match actual implementation. Grep for every flag name. (e.g., plan describes infra creation as a single command but code splits it into separate create infra and create iam commands)
Structure — component organization doesn't match actual code. Check file paths and packages. (e.g., plan describes a single controller for PSC and DNS but code has separate controllers)
Diagrams — Mermaid or sequence diagrams use names, flows, or participants that don't match actual code
The plan should reflect what IS implemented. It is not a tracker for what ISN'T.
gcp-hcp/implementation-plans/jira:create with the appropriate template from gcp-hcp/docs/).## Drift Report: {plan-name}
### Drift Items
| # | Plan Says | Code Does | Fix |
|---|-----------|-----------|-----|
| 1 | Infra creation handles WIF setup | WIF is a separate command | Update plan workflow |
### Verified (no drift)
Summary of what was checked and found to match.
### Needs Review
| # | Plan Says | Code Does | Concern |
|---|-----------|-----------|---------|
| 1 | Credentials in CPO storage component | HO creates secret directly | Breaks ownership pattern |
### Other Repos Referenced
> Ask: "This plan also references {repos}. Want me to verify against those too?"
Update this list when you find new patterns.
omitempty vs omitzero.npx claudepluginhub openshift-online/gcp-hcp --plugin gcp-hcpCompares approved plans against actual implementation to detect unimplemented items, out-of-scope changes, and design drift. Invoked via /drift-check.
Scans GitHub issues, PRs, milestones, docs, and code to flag where plans have drifted from implementation reality. Produces a cited report with prioritized reconstruction steps.
Validates code changes against original plan using git diffs and full file reads, then launches parallel subagents for quality, security, and test coverage checks. Use post-implementation or /recheck.