Skip to content

docs: add a contributor-facing capability glossary - #48

Merged
SignalLayerLabs merged 2 commits into
SignalLayerLabs:mainfrom
dchaudhari7177:docs/capability-glossary
Aug 19, 2026
Merged

docs: add a contributor-facing capability glossary#48
SignalLayerLabs merged 2 commits into
SignalLayerLabs:mainfrom
dchaudhari7177:docs/capability-glossary

Conversation

@dchaudhari7177

Copy link
Copy Markdown
Contributor

Closes #37.

Adds docs/reference/capability-glossary.md — one page an adapter PR can link to.

Acceptance criteria

  • Defines Observe, Tool Enforcement and Full Compute Enforcement — restated in substance from the overview's Integration labels section, which stays the normative source and is linked as such.
  • States that prompt instructions / advisory middleware are not enforced interception — given its own section, because it is the distinction most likely to be got wrong.
  • One current example for Codex and one for Claude Code — Codex as Tool Enforcement, Claude Code as Observe, both with the reason the codebase already gives.
  • Does not broaden any existing capability claim — see below.
  • Linked from the integration overview and contribution guidance — plus docs/index.md under Reference.

On not broadening claims

Every statement traces to an existing source. Specifically preserved:

  • Codex is Tool Enforcement, not Full Compute Enforcement, because specialized and hosted tool paths can fall outside local hook coverage.
  • Protocol v1 defines seven directives but the reference v0.2 runtime emits only allow and deny — so the page says transporting a directive is not implementing it.
  • Enforce Mode requires block_actions=True; UniversalRuntime rejects an observe-only adapter configured as enforced.
  • The Claude Code entry notes its engine-declared outcome evidence is a statement about evidence quality, not capability — richer evidence does not move an integration up a level. That seemed the likeliest misreading.

The "choosing a label" checklist resolves ties downward: under-claiming is merely conservative, over-claiming invites a user to rely on a control that fails open.

Verification

  • All 35 relative links across the four touched files resolve; both section anchors (#integration-labels, #adapter-contributions) exist.
  • Documentation-only — no runtime behavior changes.

Unrelated pre-existing failure, flagged for you

python scripts/validate_readme_pages.py fails on a clean checkout of main:

AssertionError: ## What changed after community review

README.md has no such heading (grep -c returns 0). I confirmed this by stashing my changes and re-running — it is not caused by this PR, and I have deliberately not touched README.md to fix it. Happy to open a separate issue if useful.

Observe, Tool Enforcement and Full Compute Enforcement are defined in the
integration overview's "Integration labels" section, but an adapter PR had no
single page to link to, and the reasoning a reviewer applies was spread across
the overview and CONTRIBUTING.

Add docs/reference/capability-glossary.md restating the three definitions
verbatim in substance, plus the boundaries that decide between them:

- a prompt instruction, skill, or advisory middleware is not enforced
  interception;
- transporting a protocol directive is not implementing it -- protocol v1
  defines seven, the reference v0.2 runtime emits allow and deny;
- Enforce Mode requires block_actions=True, and UniversalRuntime rejects an
  observe-only adapter configured as enforced.

Uses the two current examples the codebase already documents: Codex as Tool
Enforcement (not Full Compute Enforcement, because specialized and hosted tool
paths can fall outside local hook coverage) and Claude Code as Observe. The
Claude Code entry notes that engine-declared outcome evidence is a statement
about evidence quality, not capability, since that is the likeliest way to
misread the label.

Closes with a four-step "choosing a label" checklist that resolves ties
downward: under-claiming is conservative, over-claiming invites a user to rely
on a control that fails open.

No capability claim is broadened; every statement traces to overview.md or
CONTRIBUTING.md. Linked from docs/index.md, the integration overview and the
adapter section of CONTRIBUTING.md.

Closes SignalLayerLabs#37
…CONTRIBUTING

The glossary is only useful if an adapter PR can find it. Link it from the
three places a contributor actually looks:

- docs/index.md, under Reference;
- the "Integration labels" section of the integration overview, next to the
  normative definitions it restates;
- the adapter section of CONTRIBUTING.md, after the Full Compute Enforcement
  bar.

Refs SignalLayerLabs#37
@SignalLayerLabs
SignalLayerLabs merged commit 28eddec into SignalLayerLabs:main Aug 19, 2026
7 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

2 participants