self-reflection

Status: shipped-beta (framework-native skill, tracked in the repo) · Ships to: clone — not in any getff consumer tier arm (setup.d/lib.sh:61-63 names only the core/env+/factory skills) · Fires at: skill auto-activation on the description's trigger list — «правило», «принцип», discipline/process/meta wording, "forward check", "backward check", "self-reflection", or "any edit touching .claude/rules/, packages/core/principles/, docs/meta-factory/EXECUTION-PLAN.md, docs/meta-factory/prior-art-evaluations.md, CLAUDE.md" (.claude/skills/self-reflection/SKILL.md:3)

What it is

A recommendation-discipline gate. Its self-description: "Use when introducing or extending a rule, principle, pattern, methodology, discipline, or process change in this repository." (.claude/skills/self-reflection/SKILL.md:3). The body titles it "Self-reflection — recommendation discipline gate" (.claude/skills/self-reflection/SKILL.md:8) and scopes its authority: "Authoritative for: skill activation conditions (frontmatter description); §1.7 forward+backward checklist summary; output contract for discipline-introducing recommendations; pointers to cold references" — explicitly "NOT authoritative for: the §1.7 rule itself" (.claude/skills/self-reflection/SKILL.md:10-11). It declares a harness posture: "portable — prose self-application checklist over repo artefacts; no harness primitives" (.claude/skills/self-reflection/SKILL.md:6).

How it works

It exists because of recurrence: "Three documented occurrences of the same shape in 2026:" (.claude/skills/self-reflection/SKILL.md:15). Its output contract requires that before closing a discipline-introducing recommendation — "and in any PR description that touches discipline-bearing files" — the recommendation carries two non-empty sections "with the exact ### §1.7 heading prefix the CI gate matches" (.claude/skills/self-reflection/SKILL.md:45); a heading without the prefix does not match the gate. The skill tracks its own enforcement surface: "5 active layers as of Wave 8.1 (2026-05-12)." (.claude/skills/self-reflection/SKILL.md:108), and closes the loop on itself in "## How this skill itself complies with §1.7" (.claude/skills/self-reflection/SKILL.md:121), including a self-reflexive trigger check where the bootstrap research-patch "walks §1.7 through itself" (.claude/skills/self-reflection/SKILL.md:125).

Satellites & companions

• USES: .claude/rules/phase-research-coverage.md as the authority it points at ("the §1.7 rule itself — see .claude/rules/phase-research-coverage.md §1.7", .claude/skills/self-reflection/SKILL.md:11) — the census's B41 satellite column records this as family F (framework rules/docs).
• ADAPTS: the §1.7 forward/backward checklist into a portable prose skill with no harness primitives (.claude/skills/self-reflection/SKILL.md:6).
• ADDS: the repo's first project-internal skill — its own backward-check records that "directory empty before this commit (this is the first project-internal skill)" (.claude/skills/self-reflection/SKILL.md:124).

Anchors

• .claude/skills/self-reflection/SKILL.md:2 — name: self-reflection
• .claude/skills/self-reflection/SKILL.md:3 — description: Use when introducing or extending a rule, principle, pattern, methodology, discipline, or process change in this repository. Auto-trigger on «правило», «принцип», «дисциплина», «методология», «процесс», recommend, introduce rule, new principle, discipline change, process rule, meta, recursive, applies to itself, check own work, self-review, forward check, backward check, closing recommendation, discipline-bearing artefact, self-reflection, anti-pattern, or any edit touching .claude/rules/, packages/core/principles/, docs/meta-factory/EXECUTION-PLAN.md, docs/meta-factory/prior-art-evaluations.md, CLAUDE.md. Do NOT trigger on simple typo fixes, code edits without rule changes, or routine PR work.
• .claude/skills/self-reflection/SKILL.md:6 — <!-- @harness-posture: portable — prose self-application checklist over repo artefacts; no harness primitives -->
• .claude/skills/self-reflection/SKILL.md:8 — # Self-reflection — recommendation discipline gate
• .claude/skills/self-reflection/SKILL.md:10 — > **Authoritative for:** skill activation conditions (frontmatter description); §1.7 forward+backward checklist summary; output contract for discipline-introducing recommendations; pointers to cold references.
• .claude/skills/self-reflection/SKILL.md:11 — > **NOT authoritative for:** the §1.7 rule itself — see [.claude/rules/phase-research-coverage.md §1.7](../../rules/phase-research-coverage.md). Project goal — see [README.md#why-this-exists](../../../README.md#why-this-exists).
• .claude/skills/self-reflection/SKILL.md:15 — Three documented occurrences of the same shape in 2026:
• .claude/skills/self-reflection/SKILL.md:45 — Before closing a recommendation under skill scope — and in any PR description that touches discipline-bearing files — the recommendation must contain two non-empty sections **with the exact ### §1.7 heading prefix the CI gate matches**. [discipline-self-check.yml](../../../.github/workflows/discipline-self-check.yml) anchors its awkon^### §1.7 Forward-check appliedand^### §1.7 Backward-check applied; a heading without the §1.7 prefix does **not** match and the gate reports the section as missing.
• .claude/skills/self-reflection/SKILL.md:106 — ## §1.7 enforcement layers
• .claude/skills/self-reflection/SKILL.md:108 — 5 active layers as of Wave 8.1 (2026-05-12). Previously: 4 active layers as of Wave 7 sub-wave 7.6.c (2026-05-11). §13.23 closure shipped layer 4; §13.29 closure shipped layer 5.
• .claude/skills/self-reflection/SKILL.md:121 — ## How this skill itself complies with §1.7
• .claude/skills/self-reflection/SKILL.md:124 — - **Backward-check applied:** complete sweep of .claude/skills/— directory empty before this commit (this is the first project-internal skill); no existing entries to migrate. Exemption mechanism: skill is itself an exemption from.claude/skills/*/SKILL.mdfromprinciple 09canonical list (project-internal skills have looser authority than shippedskills/rules-as-tests/); flagged as open question for follow-up.
• .claude/skills/self-reflection/SKILL.md:125 — - **Self-reflexive trigger applied:** the [bootstrap research-patch](../../../docs/meta-factory/research-patches/2026-05-09-recommendation-skips-own-discipline.md) walks §1.7 through itself — 6/6 forward + 3/3 backward items independently catch the gap that motivated §1.7.
• setup.d/lib.sh:61-63 — GETFF_SKILLS_CORE="template-audit ai-doc rule-research rule-tests" · GETFF_SKILLS_ENV="arch night-mode orchestrator pipeline reviewer" · GETFF_SKILLS_FACTORY="dispatcher aif-doctor harvest story claude-glm-executor-handoff"

All anchors at framework pin b069c59328aa8e08671d3a19fd25efa618dfc1c5 (fetched 2026-09-11 from artyhoo/getff staging; ancestry of census pin a1337cb301 verified); .claude/skills/self-reflection/SKILL.md IS a tracked git object at that pin (read via git show b069c593:.claude/skills/self-reflection/SKILL.md), unlike the gitignored aif-* suite pages B17–B40.