Gentleman-Programming/gentle-ai / skills/rdd-advisory-transport
rdd-advisory-transport
VerifiedTrigger: reviewer transport, advisory transport, review adapter, lens prompt/schema, OpenCode reviewer plugin, Codex reviewer, ReviewProviderContract. Enforce the shared Go advisory reviewer transport contract.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 4.3 KB |
| references/issue-impact-matrix.md | markdown | 6.9 KB |
| references/shared-advisory-transport-proposal.md | markdown | 20.2 KB |
references/issue-impact-matrix.md
View on GitHub# Issue and PR impact matrix
Restored from the 267-line revision of the shared-advisory-transport proposal (read and review-verified 2026-08-06 before a later revision dropped this section). Dispositions were based on issue/PR bodies, comments, current symbols, and durable project decisions, not titles alone. Issue states move — re-verify each row against the live thread and current `main` before acting on it; a maintainer condition recorded in the issue thread always outranks this table (e.g. #2418's own closing condition governs, not its row here).
| Record | Verified evidence | Disposition and required action |
|---|---|---|
| #2207, PR #2633 | #2207 requires pre-mutation refusal for a runtime that cannot complete review. PR #2633 added the compiled policy in `review_transport_capability.go`. | **Remains valid but implementation simplifies.** Preserve pre-mutation capability admission and impossible-loop prevention. Supersede only its runtime-specific acceptance text requiring OpenCode isolation and a distinct Claude transport; admission instead requires a passing shared-contract adapter. |
| #2209 | The approved matrix separates normal SDD from immutable RDD and restricts bundled support to OpenCode, Claude Code, and eventually Codex. | **Remains fully valid.** Replace "existing bounded immutable transport" wording with "shared provider-contract conformance plus organic evidence." Keep Codex unavailable until proven; retain Pi's external-provider rule. |
| #2417, PR #2479 | Restored OpenCode shell-less provider injection; plugin logic formerly at `review-result-artifacts.ts:294-359,614-708`. | **Partially superseded.** Its user outcome, provider-owned frozen context, remains. Its implementation criteria for plugin interception, no-read/no-shell structural isolation, config checks, and dedicated OpenCode evidence path are superseded by the common advisory boundary. Keep closed; do not reopen. |
| #2418 and #2208 | #2418's maintainer decision rejects Codex hook interception and calls for provider context. #2208 requires provider binding, immutable evidence, native capture, and native E2E. | **#2418 close as obsolete ONLY per its own thread condition** (maintainer recorded an explicit closing condition there — that condition governs). Reframe #2208 as activation of the shared contract with an organic Codex adapter proof and no advertisement before that proof. Its binding, evidence, capture, and negative-E2E goals remain. |
| #2440, PR #2484 | PR #2484 binds the executing runtime identity in generated instructions and guards hard-coded identities. | **Remains fully valid and unchanged.** The shared adapter still needs the real runtime identity for capability selection and diagnostics. Do not regress the runtime-identity guards. |
| #2442 | Identifies model-authored binding JSON as the root problem; OpenCode `parseBinding` implemented field-level errors. | **Partially superseded.** Preserve the requirement that identity-bearing binding data is provider-owned and failures are diagnostic. Delete the OpenCode field-by-field parser and its prompt-rebuild advice after the Go contract owns validation. Update/close the issue against that replacement. |
| #2459, PR #2462 | PR #2462 introduced `review lens-context`; `runReviewLensContext` derives the frozen block and applies no-truncation rules. | **Remains valid but becomes the core implementation.** Fold this CLI-oriented materializer into the provider contract so every adapter consumes the same typed request. Retain its frozen evidence, full-memory assembly, and no-truncation behavior. |
| #2573, PR #2445 | #2573 keeps Codex dormant because it lacks assets; policy marks Codex eligible but `unsupported`. | **Remains valid but implementation simplifies.** Replace "missing lens assets" with "shared adapter has not passed conformance/organic proof." Keep the actionable pre-mutation refusal until the evidence exists. |
| #2624 | Its latest approved criterion mandates an isolated `opencode run` child with both global disable variables. | **Close as obsolete via the delivering PR.** This is exactly the process-isolation architecture replaced by the confirmed advisory boundary. Do not implement the child bridge, restart, dedicated session, or global environment variables. |
| #2419 | Claude live-worktree reads were admitted because prompt-only confinement was claimed as receipt-grade transport. | **Partially superseded; reframe.** Shared provider evidence/prompt eliminates the divergent empty-context path; the runtime-level structural-confinement criterion is superseded. Reframe as quality/contract-conformance: no native admission or receipt claim may depend on a model's private filesystem view. |
| #2441 | Refuter/judges had live read/MCP access; refuter lacked the common rendered transport clause. | **Partially superseded.** Preserve refuter and judge semantics; make them consume the shared provider contract. Supersede the runtime-specific no-read/no-MCP isolation posture; neither actor gains authority merely by returning output. |
| #2382 | Provider `rart1_` references cannot satisfy `capture-result --input`; the issue explicitly allows exposing schema-valid raw JSON instead. | **Remains valid but implementation simplifies.** The selected solution is the issue's second branch: adapters return raw output directly to Go admission. Do not bridge an OpenCode-only artifact reference. |
| #2566 | Claude lenses could not call `inspect-candidate`, returned incomplete, and created dead-end collection slots. | **Partially superseded.** The shared provider request removes the need for a reviewer shell or `inspect-candidate` invocation. Preserve pre-mutation admission and no-loop behavior; close only after a Claude organic run proves a completed shared-contract review without that CLI capability. |
| #2630 | Documents contract/schema drift and strict-admission diagnostics; `ReviewerResultSchema` is the native source of truth. | **Remains fully valid.** Render the shared provider contract from `ReviewerResultSchema` and native envelope metadata. Keep D2 (`--preflight` plus input) separate unless a small validation UX slice is approved. |
| #2615 | Concerns parent-orchestrator permission to execute native `review.start`, before reviewer transport begins. | **Unrelated/no impact.** Keep its normal OpenCode command-permission fix separate; do not use this work to broaden arbitrary shell permission. |
| #2645 | Proposes ReviewCore as the only semantic transition producer. | **Unrelated/no impact, architecturally compatible.** Thin adapters support its direction, but transport change must not absorb or preempt its lifecycle-authority migration. |
| #1967, #1977, #2076, #2106 | Earlier OpenCode transport incidents; policy comment records #2076's startup-binding impossibility. | **Historical implementation superseded; semantic lessons remain.** Do not reopen these closed transport incidents. Preserve frozen binding and native capture, but retire their OpenCode-only handoff, permission, restart, and parser machinery. |
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |