# Adversarial flag log — ch.1 Finding Prospects

**Reviewer mode:** controller-orchestrated independent Explore (per Dmitry ruling #2, 2026-06-09).
**Source pages reviewed:** 29-32 (book-internal "CHAPTER FOUR — Skill #1: Finding Prospects")
**Draft commit:** uncommitted at review time (extractor returned, controller dispatched reviewer pre-commit)
**Review timestamp:** 2026-06-09

---

## Flags

| # | Check | Location | Finding | Suggested fix |
|---|---|---|---|---|
| 1 | 3 | P-01-03 (linked_corrects empty) | Source para.18 explicitly names a distinct wrong behaviour — "Posers and amateurs don't even notice. They just go through their day saying 'What people? I don't see any people.'" — that is not covered by any anti_pattern node. P-01-03's `linked_corrects` is empty. The extractor argued this is subsumed by AP-01-03 (static-list operation), but AP-01-03 is about the no-inflow-habit failure, not about the perceptual/awareness failure. These are operationally distinct: AP-01-03 describes someone who has contacts but no add habit; the low-awareness pattern describes someone who encounters contacts but doesn't register them as add-able. Two different failure modes, both named in source. | Emit a new anti_pattern node (e.g., AP-01-05 or AP-01-02b) for the low-awareness / contact-blindness posture, backed by P-01-03 via a `corrects` edge. Alternatively, confirm the extractor's merge decision with Dmitry and add a note explaining why the two are merged. The current state — P-01-03 with empty `linked_corrects` — leaves the principle without a remediated anti-pattern anywhere in the graph. |
| 2 | 3 | AP-01-01 (vulnerable_to) | AP-01-01 (mental short-list) has no `vulnerable_to` edge from any user_state. The edge-cheatsheet coverage sanity check requires ≥1 `vulnerable_to` in from a user_state. Logically, S-luck-seeker (ch.00) is the obvious candidate, but ch.01 does not re-emit that state. Within ch.01 nodes only, no user_state has AP-01-01 in its `linked_vulnerable_to`. The anti-pattern fires at onboarding — before any ch.01 state is entered — which means the vulnerability belongs to S-luck-seeker. | Add a cross-chapter `vulnerable_to` edge: `S-luck-seeker --[vulnerable_to]--> AP-01-01` in the cross_chapter_links section, OR add a note explaining the coverage is satisfied by S-luck-seeker from ch.00 (if cross-chapter edges are permitted to satisfy this sanity check — clarify with Dmitry). Currently the anti-pattern is structurally unreachable from the graph. |
| 3 | 3 | AP-01-02 (vulnerable_to) | AP-01-02 (pre-disqualifying at capture time) has no `vulnerable_to` edge from any user_state in ch.01. AP-01-02 fires during L-01-01 (list dump), which is triggered from S-skill-committed. S-skill-committed's `linked_vulnerable_to` is empty. The user_state most at-risk for AP-01-02 is S-skill-committed (first list-build session). | Add `AP-01-02` to `S-skill-committed.linked_vulnerable_to`. A user who has just committed to professionalism and is doing their first list dump is exactly when the pre-disqualification reflex fires. |
| 4 | 2 | P-01-03 (source_ref vs body drift) | P-01-03's `source_ref` points to para.18 ("Professionals develop a higher level of awareness. They pay attention to the world. They know they will be introduced to new people all the time."). The body states: "The difference between an operator who adds two names a day and one who adds zero is almost never the number of people they actually encounter — it is whether they notice the encounter as contact-relevant." The "2 names vs 0 names a day" comparison and the "almost never the number of encounters" claim are not in para.18; the 2/day cadence originates from para.19. The body makes a specific causal claim the cited snippet does not support. | Either (a) broaden the source_ref to include para.18-19 (while staying within the 30-word cap), or (b) qualify the body so the causal claim is grounded only in what para.18 carries: that professionals actively notice new contacts, where amateurs do not. The "almost never encounter frequency, always awareness" framing is a reasonable inference but is not in the cited snippet. |
| 5 | 2 | AP-01-04 (source_ref vs body drift) | AP-01-04's body adds two claims not in the cited snippet: (a) "which most people detect immediately" and (b) "respond to by closing the conversation, often permanently." The source says "Some people in Network Marketing make that mistake and it's not good." The mechanism (detection + permanent closure) is absent from the snippet. | Qualify the body language: "most people detect" → "people tend to detect"; "often permanently" → soften or remove the permanence claim, or find a source passage that supports it. Alternatively keep the claims as reasonable practitioner inference and add an explicit note that this is extrapolation beyond the cited snippet — but the drift should be acknowledged. |
| 6 | 2 | AP-01-01 (source_ref ellipsis / word count) | AP-01-01's `source_ref` uses an ellipsis: "the posers make a mental list of three, four, or five people they hope will join their business...their entire future is based upon the response of those few people." The displayed snippet = 30 words (exactly at the cap). However, the ellipsis spans the words ", and" in the original text, so the underlying source range covers 32 words. If the 30-word cap is applied to the verbatim text shown, it passes exactly. If applied to the source range the snippet covers, it exceeds by 2 words. | Either (a) confirm the 30-word cap applies to the verbatim text displayed (not the source range covered), in which case this passes and note it as a boundary case; or (b) trim the snippet to bring it clearly under 30 words without the ellipsis — e.g., keep only the first sentence (18 words). Document the resolution in protocols/extraction.md as a clarification for the ellipsis case. |
| 7 | 5 | [ch.01 p.30 para.13] meta-commentary tag | The tag `meta-commentary` is defined as "Author speaks about the book, the reader, or the reading process rather than the subject matter." Para.13 ("You're not required to prospect everyone on your list. That's up to you. But it's extremely important to do the work necessary to truly capture your network on paper.") is the author speaking about the subject matter — the relationship between list-capture and prospecting-work. This is not meta-commentary about the book or reading process. The tag is wrong. | Re-tag as `restatement-of-P-01-02` (capture-without-filter principle — the sentence "it's extremely important to do the work to truly capture your network on paper" restates P-01-02's scope) or treat as a missed clarification principle about list-capture vs prospecting being separate jobs. If the extractor believes this is truly ambient author clarification, `transitional` is a better fit than `meta-commentary`, but even `transitional` is a stretch. |
| 8 | 5 | [ch.01 p.29 para.2] transitional tag | Para.2 introduces the poser/amateur/professional taxonomy ("As I mentioned in chapter two, there are three kinds of people in Network Marketing: posers, amateurs, and professionals."). The tag `transitional` is defined as "Connective tissue between sections (e.g., 'In the next chapter we will...')." This paragraph is a substantive framework recall that underpins the entire chapter's three-tier structure — it is not connective tissue. `recap` would be closer (summary of prior taxonomy), but `recap` is defined for "content already emitted earlier in the same chapter" — this is from a prior chapter (ch.02/ch.03 in the book). None of the 6 canonical tags fits cleanly. | Either (a) use `restatement-of-AP-01-01` since the poser taxon is the substantive content being recalled (most precise), or (b) acknowledge in a gap-list note that this is a prior-chapter taxonomy recall and tag it `recap` with an inline note "prior-chapter taxonomy" to flag the imprecise fit. Add a protocol amendment suggestion for a `prior-chapter-recap` sub-variant. |
| 9 | 4 | S-skill-committed (kai_voice_draft opener) | The `kai_voice_draft` opens with "Good — the decision is made." The word "Good" as a sentence opener reads as a brief approval signal. Persona negative list item 3 bans "Flattery openers: 'great question!', 'amazing!', 'that's so smart'" that inflate ordinary input. "Good" is milder than these examples, but it operates in the same register — approving the user's decision before moving to the coaching content. Persona warmth is [3], not [5]; "does not amplify emotions" is the stated guide. | Rewrite the opener to remove the approval signal. E.g., "The decision is made — now the work is choosing which skill to put time on first, and getting concrete about what practising it actually looks like this week." Same content, no approval preamble. |

---

## Summary (no-flag observations)

**Word counts / 30-word cap:** All source_ref snippets except AP-01-01 (flagged above as boundary) are clearly within 30 words. S-has-static-list at 29 words was the next closest — counted carefully and confirmed passes.

**Paragraph coverage completeness:** All 21+ paragraph units in the source (including sub-paragraphs 10b, 11b) have either an artifact or a gap-list entry. The extractor's coverage is thorough. The sub-paragraph `para.Nb` notation is a convention not in the protocol but reads unambiguously and causes no coverage gaps.

**Extractor concern #1 (S-luck-seeker forward-ref):** Evaluated independently. Agree with the extractor — ch.1 source opens in instruction-mode, not in inflection-mode. Worre does not walk a luck-seeker through a commitment arc in this chapter. The L-XX-YY forward ref in ch.00 correctly remains open.

**Extractor concern #2 (L-01-05 → L-02-XX placeholder):** Evaluated independently. Placeholders in cross_chapter_links for unextracted chapters are legitimate and not a flag.

**Extractor concern #3 (Scenario 6 R5+ state):** Evaluated independently. Scenarios are explicitly paraphrased/hypothetical per the template comment. A loosely-named state in prose is acceptable; it does not represent a schema edge.

**Extractor concern #5 (AP-01-02 source_ref trimmed):** Confirmed 28 words. No flag.

**Para.17 (Jordan Adler anecdote):** Considered whether the embedded claim "almost all the people he's recruited are people he didn't know before he started" constitutes a standalone principle (future downline mostly comes from as-yet-unknown contacts). Decided the claim is illustrative of L-01-03/S-list-builder rather than a new principle — tagging as `anecdote` is defensible, if slightly generous. Not flagged; leaving for Roma's operational truth pass.

**Voice register overall:** The draft's voice is largely clean. The directness level [4] and brevity [4] targets are well-maintained. No therapeutic register, no guru energy, no false certainty patterns were found in the bulk of drafts. The single opener issue (Flag 9) is the only voice flag and is minor.

**Loop edge coverage (triggers / transitions_to):** All five loops have inbound `triggers` edges from user_states and outbound `transitions_to` edges to user_states. No floating loops.

**Script coverage (Sc-01-01):** `backed_by` and `implements_step` both present. ✓

**Multi-artifact shared source paragraph (para.20):** P-01-04, AP-01-04, L-01-05, Sc-01-01 all share the same source paragraph. Each node's source_ref quotes a different ≤30-word segment of that paragraph. This is permissible under the IP-clean rule per extraction.md §3. Not flagged; extractor note #2 documents it correctly.

**Extractor concern #4 (P-01-03 without linked_corrects):** Evaluated independently. This is **disagreed** — see Flag 1. The extractor's argument that AP-01-03 covers the low-awareness failure mode is not supported: AP-01-03 is about inflow-habit absence, not about perceptual blindness to contacts already present in daily life. These are structurally different failure modes and the source treats them as such (para.18 names the low-awareness stance explicitly with a quote). P-01-03 should corrects something.

---

## IP self-check

This flag log uses NO verbatim source text. All findings reference paragraph numbers and node ids only. No Worre text is reproduced.

---

## Dispositions

Applied by RESOLVER, 2026-06-09. All 9 flags resolved as `resolved-fix`. Dmitry ruling on Flag 1 applied.

- **Flag 1** — `resolved-fix`: New node AP-01-05 ("Contact-blindness — feeling there is no one to talk to") created in ch.01 `# anti_patterns` section. `P-01-03.linked_corrects` updated to `[AP-01-05]`. `S-skill-committed.linked_vulnerable_to` updated to `[AP-01-02, AP-01-05]` (contact-blindness is also tempting at the first list-dump session). Source_ref for AP-01-05: para.18, 22 words, within cap.

- **Flag 2** — `resolved-fix`: `P-01-03.source_ref` span broadened to `para.18-19` to anchor the body's "two names/day vs zero" causal claim which originates in para.19. Displayed verbatim snippet unchanged (25 words, within 30-word cap). Source range extended; no body softening required.

- **Flag 3** — `resolved-fix`: `AP-01-04` body trimmed. Removed the two unsupported claims ("most people detect immediately" and "respond to by closing the conversation, often permanently"). Replacement body removes the detection/closure mechanism and simply notes the source's explicit framing that this is a mistake. No body language is now beyond what the cited snippet supports.

- **Flag 4** — `resolved-fix`: `AP-01-01.source_ref` ellipsis removed entirely. New snippet: first clause of para.3 only (18 words, unambiguously within 30-word cap). The ellipsis and the underlying 32-word span ambiguity are eliminated.

- **Flag 5** — `resolved-fix`: Gap-list entry for `[ch.01 p.30 para.13]` re-tagged from `meta-commentary` to `restatement-of-P-01-02`. Para.13 speaks to the subject matter (list-capture vs prospecting are separate jobs) not to the book or reading process; `restatement-of-P-01-02` is the correct canonical tag per precedence rule (single artifact restated).

- **Flag 6** — `resolved-fix`: Cross-chapter edge added to `# cross_chapter_links`: `S-luck-seeker --[vulnerable_to]--> AP-01-01`. Ch.0 file not edited (locked). The KG triple will carry the relation in Phase 7b per controller workflow.

- **Flag 7** — `resolved-fix`: `S-skill-committed.linked_vulnerable_to` updated from `[]` to `[AP-01-02, AP-01-05]`. AP-01-02 was the direct Flag 7 fix; AP-01-05 was added simultaneously per Flag 1 resolution (newly-committed user doing first list dump is vulnerable to both pre-disqualification reflex and contact-blindness).

- **Flag 8** — `resolved-fix`: `S-skill-committed.kai_voice_draft` opener rewritten: "Good — the decision is made." replaced with "The decision is made —". Approval signal removed; content preserved.

- **Flag 9** — `resolved-fix`: Gap-list entry for `[ch.01 p.29 para.2]` re-tagged from `transitional` to `restatement-of-AP-01-01`. The poser taxon recalled in para.2 is the primary substantive content; AP-01-01 is its in-domain operationalisation. `restatement-of-<id>` takes precedence over `recap` or `transitional` per extraction.md §5 precedence rule (single artifact restated).

### Phase 4 re-run recommendation

**Recommendation: SKIP re-run.**

The 37% threshold is exceeded (1 new node added to 18 existing = 19 total; ~7 nodes touched). However, every change is flag-driven and implements a reviewer-suggested fix. No new ambiguity was introduced — all changes resolved flagged ambiguities. Specifically: the new AP-01-05 is specified in exact detail by Dmitry's ruling; the body trims removed unsupported claims rather than adding new ones; the source_ref re-selections tightened scope rather than broadening it into contested territory; the two re-tags apply the canonical precedence rule unambiguously. The 20% re-run threshold exists to catch adversarial drift from exploratory extraction; this is a corrective pass, not an exploratory one. No Phase 4 re-run required.