Define the Action Execution Model

[SoundBlaster]

Motivation

ASP had proposal and write examples but no interoperable execution lifecycle for preview, reservation, commit, and recovery. That left mode escalation, TOCTOU, reservation conflicts, effect disclosure, and repeated compensation underspecified.

Goals

• Define one static execution mode per action and independently authorize every companion stage.
• Bind preconditions, expected effects, reservations, commits, and recovery into requests and receipts without turning evidence into authority.
• Close review proposals #19, #20, #22, #23, and #24 as one coherent Action Execution Model slice.

Changes

• Define read, dry_run, propose, reserve, commit, compensate, and revert, reciprocal companion links, required-grant closure, and canonical execution hashing.
• Add the Effect Model, preview/precondition schemas, atomic reservation lifecycle, effect-level recovery limits, revert evidence, errors, security rules, and conformance requirements.
• Mark the five selected review cards present and rebuild the standalone RFC review dashboard.

Validation

• make review-build PYTHON=/opt/homebrew/opt/python@3.10/libexec/bin/python3
• make review-check PYTHON=/opt/homebrew/opt/python@3.10/libexec/bin/python3
• Inline Python semantic validation: 785 assertions across 34 duplicate-free fenced JSON examples and 7 action declarations.
• Inline Python validation of the published JCS vectors and execution-token vector.
• Inline Python Markdown line-ending, final-newline, and trailing-whitespace check across all 4 tracked Markdown files.
• git diff --check

Boundaries and Non-Goals

• Does not add the top-level Proposal-Only Surface Mode from review card #18.
• Does not standardize action-specific semantic input normalization from review card #21.
• Does not add runtime implementation code, JSON Schema artifacts, or the broader operations-safety/event-delivery profiles.

Notes

• State-changing action grants now require an authoritative action allow-list closed over required dry-run, reservation, commit, and release dependencies; missing stages are rejected rather than added implicitly.
• Compensation and revert start a new independently authorized receipt chain and use target_receipt_hash only as causal evidence.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: fe67e29a85

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".