Checks for agent work

Components

1. Typed work result with six outcomes

   work_result_type

   Give every bounded unit of agent work exactly one typed outcome: completed with evidence, no change needed, needs input, needs scope expansion, unsupported, or budget exhausted with a checkpoint.

   Runs with

   Python 3.10 or later

   Entry points

   validate_work_result, classify_work_result, make_work_result

   Contract schemaMIT11 files, 92.2 KBruns commands

2. Provenance-aware evidence counting

   provenance_evidence_counting

   Count the independent observations behind a claim by following each report back to its root observations, so relayed, echoed and merged reports are not counted twice, and simulate non-backtracking relaying between agents.

   Runs with

   Python 3.10 or later

   Entry points

   count_independent_evidence, relay_non_backtracking

   Code moduleMIT10 files, 84.6 KBruns commands

3. Equal-cohort comparison with separate coverage

   equal_cohort_comparison

   Compare candidates only on the rows every candidate scored, and report separately how much of the declared universe each one covered.

   Runs with

   Python 3.10 or later

   Entry points

   compare_on_common_cohort

   Code moduleMIT10 files, 76.8 KBruns commands

4. Method-blind relabeling and behavioral equivalence

   behavioral_equivalence_relabeling

   Replace candidate names with seeded neutral labels kept in a separate key, and group candidates whose outputs on the same probe inputs are identical, with optional rounding.

   Runs with

   Python 3.10 or later

   Entry points

   blind_labels, equivalence_classes, blind_report

   Code moduleMIT10 files, 76.5 KBruns commands

5. Mechanism-held-out splits for synthetic task suites

   mechanism_heldout_split

   Split a synthetic task suite by generating mechanism so that no template, rule family or simulator appears on both sides, build balanced mechanism-level folds, and detect leakage in any given split.

   Runs with

   Python 3.10 or later

   Entry points

   split_by_mechanism, group_k_folds, leakage

   Code moduleMIT10 files, 79.1 KBruns commands

6. Silent concept-shift twin datasets

   silent_concept_shift_twins

   Generate two datasets whose inputs and targets have exactly the same marginal distributions while the input-target relationship is reversed, and compare them with input-only, target-only and joint statistics.

   Runs with

   Python 3.10 or later

   Entry points

   generate_twins, input_only_statistics, target_only_statistics, joint_statistics, silent_shift_report

   Code moduleMIT10 files, 79.5 KBruns commands

7. Validate an append-only ledger linking each artifact version to what changed

   baltor_evidence_learning_ledger

   Validate an append-only ledger linking each artifact version to what changed, why, what failed, what improved and the lesson learned, refusing any entry that removes, edits or replaces an original instead of sitting beside it.

   Command

   python3 -B scripts/tool.py < examples/example.input.json

   Input fields

   previous_entries, entries

   SkillMIT10 files, 53.0 KBreads files, runs commands

8. Compare consecutive versions of one artifact with the defects found

   baltor_iterative_improvement_record

   Compare consecutive versions of one artifact with the defects found, the fix or reason for each and a severity-ordered before-and-after verdict, refusing a version that drops a recorded defect without a fix or a reason or claims a better verdict than the counts show.

   Command

   python3 -B scripts/tool.py < examples/example.input.json

   Input fields

   artifact_id, versions, resolutions, claimed_verdicts

   SkillMIT10 files, 53.0 KBreads files, runs commands

9. Compare two counted proportions with exact arithmetic and explicit denominators

   baltor_percentage_comparison

   Compare two counted proportions with exact arithmetic and explicit denominators.

   Files

   SKILL.md

   Placed for

   OpenCode

   SkillMIT12 files, 33.3 KBreads files, runs commands

10. Check technical writing in Markdown or plain text against twelve measurable clarity rules

    baltor_plain_technical_writing_check

    Check technical writing in Markdown or plain text against twelve measurable clarity rules and report each finding with its line, column, paragraph and sentence.

    Command

    python3 -B scripts/tool.py --check draft.md

    Input fields

    text, format, limits, rules, known_abbreviations, source_name

    SkillMIT11 files, 99.3 KBreads files, runs commands