Overview
A stewardship instrument for maintained commitments: systems, obligations, and relationships to past work. It reads a declared file of commitments against a versioned registry and returns a fail-closed declaration-coverage verdict (MAINTAINED, NEEDS_ATTENTION, STALE, OUTSIDE_SCOPE) over the care signals each commitment requires. It measures whether the care a steward declared was fresh at the review date — never whether the care happened; a MAINTAINED verdict is not a warranty, SLA, availability guarantee, or proof of maintenance, and it never authorizes an action. License: Apache-2.0. See LICENSE in the repository.
Overview source: Curated paper metadata.
Methods and contributions
Read the source for the full argument, qualifications, and evidence.
Findings and contributions
- An empty scan set (no matching commitment or unreadable file) returns OUTSIDE_SCOPE with notes and never yields MAINTAINED.
- The stated staleness-monotonicity property holds that the stale set is always a subset of the missing set.
- The verdict is a declaration-coverage statement, explicitly not a warranty, SLA, availability guarantee or proof of maintenance.
- The author notes the instrument only reads declarations and does not inspect repositories, run checks or contact anyone; the freshness window is a policy constant, not empirical.
Methods
- Versioned registry of twelve commitments with SHA-256 digest pinning — Commitments are frozen records (id, title, wire, tags, required care signals, kind) whose sorted canonical form is hashed and pinned in every reading.
- Staged, fail-closed intake normalization of declared files — Declared files are normalized; malformed tags or undated signals are set aside as intake notes rather than causing crashes or invented values.
- Freshness-window evaluation into MAINTAINED / NEEDS_ATTENTION / STALE — Required signals of tag-matched commitments are compared with signals declared within a freshness window (180 days by default) to assign a verdict.
- Formal definition of readings, commitments and fresh sets — A reading is formalised as a tuple and properties (fail-closed, staleness monotonicity, digest pinning) are stated.
- Worked examples: maintained quarter, stale repository, unmatched scan — Three illustrative declared files show the verdicts the instrument returns; these are examples, not an empirical evaluation.
Key concepts
- stewardship
- maintenance
- line set
- accountability
Summary sources: Paper metadata and evidence · Learning guide · Extracted source text.
PDF downloads
Archived files available directly from this site.
- blue_line_combined.pdf PDF · 0.08 MiB · extracted-text source
Citation
Citation metadata follows the unified bibliography.
Catalog details and resources
Related in Active Inference
Other catalogued works in the same domain.