v0.2.7: simplify instructions and clarify installation

v0.2.7: simplify instructions and clarify installation

Experimental prerelease, November 7, 2026.
Bundle: context-docs 0.2.7, adopt-context-docs 0.1.16 and unchanged
standard 0.1.1. Previous bundle: v0.2.6.
Publication and installation are separate actions. The owner separately authorized website deployment; installed upgrades remain separate.

What changes

The core skill consolidates the canonical-location rule and removes an incidental
ASD-STE100 example. Adoption shortens repeated coordination, storage and setup
instructions while continuing to require the sibling core skill. The release
retains the existing clarity, preservation, audience, scope, progress, logging,
shared-file coordination and publication safeguards. It adds no background service,
file synchronization, storage migration or new runtime dependency.

The accompanying documentation:

• Pins everyday installations to published v0.2.7 (core 0.2.7 and adoption 0.1.16),
  verifies the expected versions and labels development-source installation separately.
• Links newer feedback and focused studies from the evidence overview,
  preserving each record's versions, method, outcomes and limitations.
• Reminds users to transfer shared context alongside code within an already-authorized
  Git workflow and check the destination checkout before resuming. Ignored local-only
  context needs an existing private transfer or backup procedure.
• Keeps public contributor context focused on the project and its readers.
  Actual model provenance, attribution and frozen evidence remain intact.

Evidence and qualifications

The simplification candidate check
contains three fresh Codex sessions and a separate three-session Claude extension
using this candidate's unchanged package bytes. Focused refresh/preservation,
unchanged-repeat and re-adoption file checks pass on a reused, guided synthetic fixture.
Codex reviewed the skill implementation; the subjects were not independent graders.

The Codex re-adoption report's automatic-loading confirmation remains unverified.
Claude startup/import events support loading in its sessions only. A post-run
path-alias parser correction is retained without rerunning the subjects or
changing acceptance rules. These checks do not cover fresh adoption or implicit
selection and establish no general reliability, privacy, comprehension or effort benefit.
The earlier optional skill validators could not start without PyYAML; no dependency
was installed for those attempts. Repository metadata/link/version checks are separate.

The anonymous feedback
includes four individual responses and one team response, all reporting usefulness
and intent to continue. The owner separately confirmed human-observed usability
and cross-machine use. Exact participant package/client versions, session traces,
measured comparative effort and use of the frozen scored trial remain unestablished;
these reports are not v0.2.7-specific validation.

The Git-transfer check
used released core 0.2.6/adoption 0.1.15 on one host, not separate physical machines.
The stale-note check
used source core 0.2.7 and corrected one stale dependency note on explicit refresh;
it did not reproduce the participant's reported miss or require another skill fix.
Neither adds a comparator or independently scored user outcome. Original evidence
and earlier release qualifications remain unchanged.

Existing projects and upgrades

Publication and installation are separate actions. Existing installations keep their
current guidance until separately upgraded. Preserve customizations and both complete
sibling packages; follow the upgrade guide.
Keep context locations, layouts, decisions and rationale, maintenance rules,
adoption dates, local-only boundaries and logging settings. Publication
does not authorize moving records or rewriting historical evidence. Audits remain read-only.

Preparation and publication

The local preparation record retains its original unpublished
scope and records source identities,
checks and preservation. Current local tests are distinct from the earlier model
sessions; no new native evaluation or counterpart model review ran during preparation.
Validation of the final local commit is retained in private operator notes with its
exact commit ID. Publication runs applicable commit checks and the
release checklist, including isolated
skills.sh verification. Website deployment and installed upgrades remain separate actions;
release publication alone does not establish live-site changes or native skill behavior.

Publication and discovery verification

Published the experimental prerelease
from 00c4b51 after both release-commit Actions passed. The
publication record owns exact source, CI and verification.
The separately authorized website deployment uses the same source. All ten uploaded
files match its repository build. Six live non-HTML resources match byte-for-byte;
each of four HTML pages differs only by one Cloudflare analytics beacon injected
outside the repository. The owner chose to retain and record that difference.
The remaining HTML matches exactly; desktop/mobile checks pass on all four routes.

One filesystem-confined skills@1.7.4 install into a temporary project copied all
16 release package files exactly, including core 0.2.7/adoption 0.1.16. All 403
protected installed entries stayed unchanged; the original telemetry response
was HTTP 200. Both skills.sh full rendered skill bodies still matched v0.2.6
after the bounded processing/cache interval. Public page refresh remains unverified.
No second install or telemetry replay was performed. Installation success and
telemetry acceptance do not establish page-content update, supporting-file
snapshots, counted installs, audit reruns or future refresh reliability.