From 4fa411f13ca146f50ac96b3007689cd62fe27366 Mon Sep 17 00:00:00 2001 From: Kjell Tore Guttormsen Date: Tue, 23 Jun 2026 18:23:04 +0200 Subject: [PATCH] =?UTF-8?q?feat(linkedin-studio):=20SB-S3a=20=E2=80=94=20w?= =?UTF-8?q?ire=20strategy-advisor=20as=20first=20brain/profile.md=20reader?= =?UTF-8?q?=20[skip-docs]?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The second brain now feeds generation: strategy-advisor reads brain/profile.md as evidence-to-test (anti-sycophancy default, graceful absence on fresh installs). First end-to-end proof of capture → consolidate → read-back-into-generation. - agents/strategy-advisor.md: brain/profile.md added to Step 0 Load Context + a consumption subsection (Static/Dynamic layers, evidence_count/last_seen weighting, anti-sycophancy counter-pressure, silent degrade when the file is absent). - scripts/test-runner.sh: new Section 16d (Brain Profile Reader) — 2 UNCONDITIONAL checks (non-vacuity self-test with legacy-path/lowercase decoy + exact-literal wiring grep on strategy-advisor.md), ASSERT_BASELINE_FLOOR 78→80, header enumeration extended. Gate 93→95/0/0; brain floor 82 untouched. - docs/second-brain/consolidation-loop.md: reconciled the stale "no reader yet" honest-limit to the true partial state (one reader; broader consumption deferred). TDD: RED (only the wiring grep failed, exit 1) → GREEN (95/0/0). READ-only — the brain consolidate --apply --confirm gate stays the sole writer. SC4 (graceful absence) verified by inspection; SC5 (read-back) deferred to a reloaded session (honest — agent prompts load at session start). S3a scope held: one reader, no parser, no id-threading. Co-Authored-By: Claude Opus 4.8 (1M context) Claude-Session: https://claude.ai/code/session_01RigJBiRFNtFZKCz21qNbQ4 --- agents/strategy-advisor.md | 14 ++++++ docs/second-brain/consolidation-loop.md | 6 ++- scripts/test-runner.sh | 62 +++++++++++++++++++++++-- 3 files changed, 76 insertions(+), 6 deletions(-) diff --git a/agents/strategy-advisor.md b/agents/strategy-advisor.md index 7ae5d31..3f0dee0 100644 --- a/agents/strategy-advisor.md +++ b/agents/strategy-advisor.md @@ -44,10 +44,24 @@ ${LINKEDIN_STUDIO_DATA:-$HOME/.claude/linkedin-studio}/audience-insights/engagem ${LINKEDIN_STUDIO_DATA:-$HOME/.claude/linkedin-studio}/examples/high-engagement-posts.md → proven patterns from top posts ${CLAUDE_PLUGIN_ROOT}/references/trajectory-strategy-adjustments.md → trajectory-to-action mappings ~/.claude/linkedin-studio.local.md → user state + posting history +${LINKEDIN_STUDIO_DATA:-$HOME/.claude/linkedin-studio}/brain/profile.md → evolving second-brain profile (consolidated, evidence-weighted facts about the user) ``` Use demographics data to compare the user's **intended** audience vs **actual** engagers when making strategic recommendations. +### Consuming the evolving profile (brain/profile.md) + +The second-brain profile is a consolidated, evidence-weighted picture of the user, grown over time by the brain motor (mutated ONLY via `brain consolidate --apply --confirm` — you never write it). It has two layers: + +- **`## Static`** — stable, high-confidence facts about the user. +- **`## Dynamic`** — emerging facts still accruing evidence. + +Each fact line carries `evidence_count` and `last_seen`. Weight by strength and recency: a higher `evidence_count` and a more recent `last_seen` are a stronger, fresher signal than a thin or stale one. + +**Anti-sycophancy (binding):** Treat every profile fact as **evidence to TEST**, not flattery — counter-pressure it against the analytics and state each time you use it (e.g. *the profile says you lean toward X, but your recent imports show Y — test that before recommending*). The profile informs your recommendations; it never dictates them and never flatters the user. + +**Graceful absence:** If `brain/profile.md` is missing or empty (fresh installs have no brain yet), proceed silently on the other context sources — no error, and no note about a missing profile. + ### New Creator Advantage Detection From the state file, extract `first_post_date`. Calculate the creator window status: diff --git a/docs/second-brain/consolidation-loop.md b/docs/second-brain/consolidation-loop.md index f88b852..6aae33d 100644 --- a/docs/second-brain/consolidation-loop.md +++ b/docs/second-brain/consolidation-loop.md @@ -64,8 +64,10 @@ Defaults: `promoteThreshold = 3`, `decayDays = 90` (operator-confirmed). - **The loop's value depends on the session's extraction.** The engine only guarantees threshold/conflict/ decay/provenance mechanics. Garbage candidates → a garbage diff. The operator gate + candidate-shape validation catch shape errors, not insight quality. -- **`brain/profile.md` has no reader yet.** S2 evolves the profile; wiring content agents/commands to - *consume* it is SB-S3. The value is deferred: the profile compounds now so S3's reader inherits rich data. +- **`brain/profile.md` has one reader as of SB-S3a.** S2 evolved the profile motor-only; SB-S3a wired the + first consumer — `strategy-advisor` reads it as *evidence-to-test* (anti-sycophancy: counter-pressured, never + parroted). Broader consumption (more content agents, a hook-level prompt digest) remains later S3 work, and + the profile is still mutated ONLY via `brain consolidate --apply --confirm`. - **No supersede / no auto-demotion.** Conflicts keep both; stale facts are flagged, never auto-removed — the operator (or S3) reconciles. Conflict alt facts persist until then. - **No AI at session-start.** The nudge is a deterministic file-count + sidecar read; the consolidation diff --git a/scripts/test-runner.sh b/scripts/test-runner.sh index 50ba718..4b08af8 100755 --- a/scripts/test-runner.sh +++ b/scripts/test-runner.sh @@ -30,8 +30,10 @@ # findings through it — KTG-only, skipped for an adopter shipping no deps) in Section # 16; the trend-spotter de-niche guard (B-S1: agents/trend-spotter.md names no # hardcoded vendor/sector beat — the domain comes from the user's pillars at runtime, -# never baked into the agent — with a non-vacuity self-test) in Section 17; the -# assertion-count anti-erosion floor (SC6) in Section 18. All are live below +# never baked into the agent — with a non-vacuity self-test) in Section 17; the brain +# profile-reader guard (SB-S3a: strategy-advisor names brain/profile.md AND carries the +# anti-sycophancy literal 'evidence to TEST', with a non-vacuity self-test) in Section +# 16d; the assertion-count anti-erosion floor (SC6) in Section 18. All are live below # (Sections 8–18). # # Usage: bash scripts/test-runner.sh @@ -779,6 +781,56 @@ fi echo "" +# --- Section 16d: Brain Profile Reader (SB-S3a) --- +echo "--- Brain Profile Reader ---" + +# SB-S3a wires the FIRST reader of the evolving second-brain profile: strategy-advisor +# must (a) name brain/profile.md as a context source AND (b) carry the anti-sycophancy +# literal 'evidence to TEST' (architecture.md:58 — the profile informs, it never dictates +# or flatters; every fact is counter-pressured against analytics/state). Both literals +# are required, grepped EXACT with grep -F: a loose 'profile' pattern would match the +# legacy flat profile/user-profile.md vacuously, and a loose lowercase 'evidence' would +# match generic prose. Non-vacuity self-test mirrors Sections 8/13/16c/17: a probe counts +# as "wired" iff it carries BOTH exact literals; probes missing either (incl. a decoy +# naming the legacy path + lowercase "evidence") must NOT count as wired. The wiring is +# gate-enforced; the agent's runtime OBEDIENCE is verified manually (SC4/SC5 at land — +# agent-prompt behaviour is not unit-testable; the brief states this honestly). +READER_LIT='brain/profile.md' +ANTISYC_LIT='evidence to TEST' + +reader_wired() { # $1 = text; wired iff BOTH literals present (echo twice — grep consumes stdin) + echo "$1" | grep -qF "$READER_LIT" && echo "$1" | grep -qF "$ANTISYC_LIT" +} + +READER_SELFTEST_OK=1 +if ! reader_wired "reads brain/profile.md as evidence to TEST"; then + READER_SELFTEST_OK=0; echo " non-vacuity FAIL: a fully-wired probe was not detected" +fi +while IFS= read -r probe; do + [ -z "$probe" ] && continue + if reader_wired "$probe"; then + READER_SELFTEST_OK=0; echo " false-positive FAIL: under-wired probe accepted -> $probe" + fi +done <<'NEGATIVE16D' +reads brain/profile.md but never frames how to weigh it +treats facts as evidence to TEST but names no source file +reads profile/user-profile.md and tests the evidence +NEGATIVE16D +if [ "$READER_SELFTEST_OK" -eq 1 ]; then + pass "profile-reader self-test: full wiring detected; 3 under-wired forms rejected (incl. legacy-path + lowercase 'evidence' decoy)" +else + fail "profile-reader self-test failed — the reader-wiring lint is vacuous or over-eager" +fi + +SA="agents/strategy-advisor.md" +if grep -qF "$READER_LIT" "$SA" && grep -qF "$ANTISYC_LIT" "$SA"; then + pass "strategy-advisor.md wired to brain profile reader (names '$READER_LIT', frames '$ANTISYC_LIT')" +else + fail "strategy-advisor.md missing brain profile reader wiring — needs both '$READER_LIT' and '$ANTISYC_LIT'" +fi + +echo "" + # --- Section 17: De-Niche Guard (B-S1 + B-S2) --- echo "--- De-Niche Guard ---" @@ -851,12 +903,14 @@ echo "" # assertion. Pin the total pass()+fail() invocations as a monotonic floor; the count # may only grow (brief-reviewer assumption 3). History: 74 pre-M0; +1 for the SB-S0 # brain-suite floor (Section 16b) = 75; +3 for SB-S1's three UNCONDITIONAL Section-16c -# checks (published-only self-test + voice-trainer grep + contract-doc grep) = 78. +# checks (published-only self-test + voice-trainer grep + contract-doc grep) = 78; +# +2 for SB-S3a's two UNCONDITIONAL Section-16d checks (profile-reader self-test + +# strategy-advisor wiring grep) = 80. # NB: the floor tracks the deps-absent MINIMUM (conditional TS suites warn-skip and drop # the count), so it is bumped only by UNCONDITIONAL new checks — NOT pinned to the # deps-present TOTAL_CHECKS (that would zero the warn-skip margin and false-fail a fresh # clone). Runs last so TOTAL_CHECKS sees every prior check. -ASSERT_BASELINE_FLOOR=78 +ASSERT_BASELINE_FLOOR=80 TOTAL_CHECKS=$((PASS + FAIL)) if [ "$TOTAL_CHECKS" -ge "$ASSERT_BASELINE_FLOOR" ]; then pass "assertion-count anti-erosion: $TOTAL_CHECKS checks >= baseline floor $ASSERT_BASELINE_FLOOR"