Repository navigation
feat(research): add strict digital metrics protocol - #46
erinepshovel-code wants to merge 25 commits into
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3ed3f06916
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 22d7ec9273
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 27ac5e0186
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 4bc32e4100
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ba7e539670
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8a4a4805f8
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".
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e0c054bee4
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".
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 96b2d40f58
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".
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f37b1774ac
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".
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 167c829cf3
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".
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 18f3aa1e65
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b736309726
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: bfc1feb13a
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 9229fe57f5
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ae41ff2c15
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7d886cd4b1
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".
erinepshovel-code
left a comment
erinepshovel-code
left a comment
There was a problem hiding this comment.
Publishing accumulated exact-head fix receipts; current merge remains gated on a fresh review of 88079f5.
|
@codex review exact head 88079f557f21f199e69d40f6fff88f608d94485d |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 88079f5522
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2b1b0825d9
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f6cf01a936
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 841334406a
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".
| report_fd = os.dup(1) | ||
| report_write = os.write | ||
| report_close = os.close | ||
| shutdown_marker = base64.b64decode(payload.pop("shutdown_marker")) |
There was a problem hiding this comment.
When witness code walks sys._getframe().f_back to _run_isolated_witness, it can read report_fd and shutdown_marker from that frame's locals, write a forged clean execution report and marker with os.write, then call os._exit(0). Fresh evidence after moving these values from globals to function locals is that, at exact head, this still makes _run_unittest_witnesses() return (True, '', 1) without executing an admitted failing test, allowing its contracts to receive proof credit; construct and authenticate the report outside the interpreter that executes witness code.
AGENTS.md reference: AGENTS.md:L96-L102
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Confirmed at current head 28647b5 and intentionally left unresolved. Arbitrary Python witness code cannot provide unforgeable evidence about its own execution while sharing the interpreter or any reachable result channel; moving the descriptor, token, or callback only relocates the bypass. A defensible fix requires a separately trusted execution monitor/sandbox or replacing executable Python witnesses with a restricted declarative witness language. Merge is held pending that architectural choice.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 28647b5b29
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".
| changed_result_callbacks = sorted( | ||
| name | ||
| for name, callback in trusted_result_callbacks.items() | ||
| if AppendOnlyWitnessResult.__dict__.get(name) is not callback | ||
| ) |
There was a problem hiding this comment.
When a witness replaces the result class's addError and addFailure callbacks with no-ops, registers a failing cleanup, and then restores both callbacks in later LIFO cleanups, the failure is swallowed and this post-run identity comparison sees the restored originals. At the reviewed tree, that fixture makes _run_unittest_witnesses() return a clean one-test result despite the cleanup raising AssertionError; fresh evidence beyond the previously fixed class-level replacement path is that restoring the callbacks before validation evades the new check. Protect these callbacks throughout cleanup execution or record outcomes through a channel witness code cannot override.
AGENTS.md reference: AGENTS.md:L96-L102
Useful? React with 👍 / 👎.
| source = re.sub(r"<!--.*?-->", "", source, flags=re.DOTALL) | ||
| source = re.sub(r"<!--.*\Z", "", source, flags=re.DOTALL) |
There was a problem hiding this comment.
When the complete ### Digital metrics exact projection section is wrapped in <pre>...</pre>, Markdown treats its headings and tables as raw preformatted text, but this preprocessing retains every inner line and the checker still accepts them as the rendered projection. I reproduced check_digital_metrics_projection() returning no findings for that mutation; fresh evidence after the comment and fence hardening is that raw HTML blocks provide another way to hide the required human authority projection while the structural gate passes. Parse rendered Markdown semantics or exclude raw HTML blocks before section discovery. .agents/skills/stack-update/SKILL.mdL89-L101
Useful? React with 👍 / 👎.
| for field in ("commit", "authority", "relation"): | ||
| if manifest_entry.get(field) != graph_entry.get(field): | ||
| error(findings, "digital_metrics.projection", f"{repository}: {field} differs from work graph") |
There was a problem hiding this comment.
When a participant's authority is changed consistently in the work graph, machine manifest, and human authority table, this loop checks only projection equality and accepts the new claim. I reproduced a self-digested graph and manifest that assign EDCM semantic definitions, geometry, and proof status while all transfer flags remain false, and the complete structural checks inspected here return no findings; fresh evidence after the fixed-boundary hardening is that the authority strings themselves remain unconstrained. Require the v0 participant authority split so coherent edits cannot transfer METAPAT or UCNS ownership to EDCM.
AGENTS.md reference: AGENTS.md:L23-L25
Useful? React with 👍 / 👎.
| def load_work_graph(path: Path) -> dict[str, Any]: | ||
| value = json.loads(path.read_text(encoding="utf-8")) | ||
| expected = {"schema", "version", "participants", "boundaries", "work_graph_sha256"} | ||
| if not isinstance(value, dict) or set(value) != expected: |
There was a problem hiding this comment.
When WORK_GRAPH.json repeats a member such as commit, Python's json.loads silently keeps the last value, so this validator and the stack checker validate and digest only the collapsed mapping rather than rejecting the ambiguous source. I reproduced a participant containing an all-zero commit followed by the currently pinned commit being accepted with the unchanged work-graph digest; generation and replay can therefore attest the same receipt even though the repository-owned authority record encodes two source identities and another JSON consumer may select the first. Parse with duplicate detection before validating or hashing the graph. .agents/skills/stack-update/SKILL.mdL89-L97
Useful? React with 👍 / 👎.
1 participant
Summary
Validation
Standing
INCUBATING / TEST-BACKED STACK RESEARCH. This establishes strict representation, structural observation, and replay only. EDCM projection selection, empirical calibration, producer signatures, and an independently implemented verifier remain unresolved.
hmmm: the first irreducible assumption is the selection of an EDCM projection from an exact structural observation to a named construct; the next resolving experiment must preregister that projection and evaluate a blind held-out set against matched controls.