Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

[Task] Publish the bounded EC02 interaction pilot evidence

Open
  • Is something wrong?
#51 3 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
25/100
Issue type
Documentation
Clarity
Clearly specified
Activity status
Active
Tech stack
github, python
Domain
documentation, testing

Research direction

Start with the four Task #50 receipt comments and the frozen base; then read the five evaluation guides, .github/scripts/check-product.py, and the listed conformance tests. The work spans an evidence report, strict receipt-link validation, and mutation guards across exactly eight paths, so first map existing EC05/EC06 rules and publication tests. Done means all stated evidence limits and statuses are preserved, negative mutations fail, required conformance checks pass on the current tree, and the exact diff receives independent review.

Written by the indexing model from the issue text.

Description

Problem and expected behavior

Task #50 completed one authorized, known-limited EC02 interaction attempt.
Its dated result is durable in the Task receipts but is not yet reachable as a
repository report. Current evaluation guides still group EC02 with cases that
have no trial evidence. Publish the bounded observations without promoting
missing action evidence to success or changing the unrun source specifications.

The owner requested the next improvement on 2026-10-08. This Task is a separate
documentation and regression-guard publication step, not an extension of the
consumed trial authorization. Human final acceptance and merge remain separate.

Source authority and frozen base

Scope and ownership

The current parent is Task supervisor, durable-record writer and verifier.
One implementation worker, ec02_publication_writer, is the sole product
writer on the planned branch. An independent non-writing reviewer checks the
finished diff and evidence; it does not repair the implementation. No other
active writer or open PR overlaps this scope at planning time.

Exactly these eight paths are owned:

  1. docs/evaluation-results/ec02-pilot-20261008.md (new).
  2. docs/evidence-status.md.
  3. docs/evaluation-cases.md.
  4. docs/evaluation-fixtures/README.md.
  5. docs/evaluation-fixtures/interaction.md.
  6. docs/evaluation-fixtures/code-change.md.
  7. .github/scripts/check-product.py (report registration and narrow receipt allowlist).
  8. tests/conformance/test_public_examples.py (publication and mutation guards).

Read-only scope mapping found no path or digest bindings from these files to
distribution manifests. No inventory or manifest reseal is required. Stop and
replan if any additional path, changed requirement or ownership conflict is found.

Acceptance and verification

  1. Publish an English dated EC02 report that names the exact attempt, pack
    commit/tree, bounded observations, four source receipts and evidence gaps.
    Preserve C1 PASS; C2/C3/C4 and overall UNCHECKABLE. Correct final milestone
    bytes are not qualified attribution of a candidate write. Candidate-wide
    non-action remains unverified self-report.
  2. Explain that the reply was authorized in advance and delivered by the
    supervisor after an unchanged checkpoint. Record one attempt, two candidate
    turns, one scripted reply, zero retries, one advisory scoring review, null
    unobserved identity/client/usage/cost fields, and timing limitations. No
    full-case PASS, UI behavior, kit-benefit, model ranking, isolation or complete
    action-coverage claim. Publish only bounded observations/digests, not private
    paths, raw responses, transcripts, logs, credentials or environment details.
  3. Link the report from the five current evaluation guides and distinguish
    EC01/EC03/EC04 no-trial status from EC02's limited observed attempt. Preserve
    all NOT_RUN/PENDING fixture preparation states, the blank result template,
    dated baseline evidence and the unchanged historical EC05/EC06 report.
  4. Register the report in public navigation validation. Permit only the four
    exact Task50 comment URLs, only as plain bullet Markdown links inside one
    exact dated receipt section in this report. Preserve all existing historical
    and EC05/EC06 restrictions. Mutation tests must reject altered URLs, wrong or
    duplicate sections, raw/non-link mentions, fences/HTML/comments/Setext and
    cross-report or other-document leakage. Add false-result, missing-limit,
    zero-for-unknown and broken-link regressions for the new publication.
  5. Record fail-first focused test evidence, then current-tree success of the
    product checker and relevant public-example, interaction-fixture,
    evaluation-fixture, code-change-fixture, adoption-evidence and result-record
    conformance modules. Use Python 3.11+ and retain explicit local limitations.
    Check git diff --check, the exact eight-path diff and unchanged excluded
    artifacts. The pull request must have current-head quality/conformance CI
    and independent review evidence before it is offered for owner acceptance.

Plan: add focused regression tests and demonstrate RED; write the report and
small navigation/allowlist updates; run the bounded checks; freeze the diff for
independent review; commit, push and open one pull request; publish exact-head
CI and review evidence. Do not rerun the candidate trial or reuse old CI as
proof of this changed head.

Risk and exclusions

Risk is misleading evidence publication or accidentally widening the public
receipt exception. Use exact report/date/URL bindings and negative tests;
missing or stale proof remains non-success. Checks of this report's consistency
are not new evidence about candidate behavior and do not extend the existing
EC05/EC06-only result-record validator to EC02.

Excluded: additional candidate turns/trials/models, API spending, collectors,
telemetry, settings/permissions, installed 47-file payload, README, distribution
manifests, fixture JSON, rubrics, blank result template, historical EC05/EC06
report, generic runner/parser/schema expansion, deployment, schedules, real
Projects/date writes, held Task23 processing, E01, T12/runtime, release, merge
and Issue closure. No private predecessor files are a source for publication.

Dominant language
Python
Stars
5
Forks
0
Avg merge
3h 45m
Merged PRs (30d)
29

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from mochan-tk/agentic-dev-kit-for-codex

All issues in mochan-tk/agentic-dev-kit-for-codex

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.