Activity Boost Package — STAGED, awaiting 3-day test verdict

This package is prepared but not executed. It will be applied only if the 3-day test (Tuesday May 26 through Thursday May 28) returns soft-pass or fail per the rubric in §4 below. If the test returns pass, this package stays staged for a longer-window decision.

The package contains four changes — two scheduled-task prompt edits, one cron change, one new section in this same file's parent (Backlog.md). It does not modify soul files, framework files (08-Frameworks/*), the vault folder structure, file naming conventions, the verdict enum, the MoS bands, or any artifact template.

1. Test rubric — what the 3 days have to produce

Tuesday May 26 (clean open-market day post-Memorial-Day), Wednesday May 27, Thursday May 28 = three real trading days for the current pipeline to demonstrate it works.

Pass — package stays staged, current pipeline keeps running:

≥3 thesis bundles produced across the 3 days (any combination of triggered + Q2 fallback)
≥3 watchlist additions
At least one continue first-read decision recorded
At least one verdict landing as buy OR pass-with-trigger with a near-current trigger (within 15% of price)

Soft pass — execute package as Tier 2:

1-2 thesis bundles, OR
0-2 watchlist additions despite first-reads being written, OR
Most first-reads decide pass at scan stage (criterion #1 rejection), OR
Builder exits "no theses to build today" on ≥1 of the 3 days

Fail — execute package immediately, no further deliberation:

0 thesis bundles produced
0 watchlist additions
0 continue first-reads
Builder exits "no theses to build today" on ≥2 of the 3 days

2. Throughput expectations after the package executes

Metric	Current	Post-package
Thesis bundles / week	~1-2	~7
Watchlist names after 30 days	1-3	15-25
First-reads / week	1-3	4-7
Builder run with "nothing to do"	most weekdays	<1/week

The capital discipline gate is unchanged — buy verdicts still require the same standard. The change increases the probability of a trigger firing per unit time by widening the bench, not by lowering the bar.

3. The four changes — exact specifications

Change 1: Thesis-builder fallback queues

File: C:\Users\kenne\OneDrive\Documents\Claude\Scheduled\alphasteve-thesis-builder\SKILL.md

Edit location: Section "### Step 1 — pick the work item" (around line 32-33 of current file)

Current text:

### Step 1 — pick the work item
Open `14-Optimization/Backlog.md`. Find the oldest P1 item of the form "Build full thesis on TICKER — escalated from first-read YYYY-MM-DD." Skip items marked "blocked" or "in progress" by another run. If none exist, write one line in today's run log: "No theses to build today" and exit cleanly.

Replace with:

### Step 1 — pick the work item

Try queues in this priority order. Stop at the first queue that produces a pickable item.

**Q1 — Triggered (from scan continues, highest priority):** Open `14-Optimization/Backlog.md`. Find the oldest P1 item of the form "Build full thesis on TICKER — escalated from first-read YYYY-MM-DD." Skip items marked "blocked" or "in progress." If found, work is a triggered thesis — verdict spans full range (buy / pwt / pass / avoid based on the work).

**Q2 — Shelved first-reads (graduate to full thesis):** If Q1 is empty, scan `09-Theses/<TICKER>/first-read-YYYY-MM-DD.md` files where frontmatter `decision: shelve-with-trigger` AND no corresponding instance folder `<TICKER>/<TICKER> - <MMMYYYY>/` exists. Pick the oldest by `first_read_date`. Work is a coverage graduation — default verdict: `pass-with-trigger` matching the first-read's shelve_trigger value.

**Q3 — Coverage queue (user/agent-curated):** If Q1 and Q2 are empty, read `14-Optimization/Backlog.md` for the `## Coverage queue` section. Pick the top entry (ticker row with optional source note). Work is a coverage initiation — default verdict bias: `pass-with-trigger` unless the bundle work reveals a buy or structural avoid.

**Exit:** If all three queues are empty, write "No theses to build today (Q1/Q2/Q3 all empty)" in the run log and exit cleanly. This should be rare.

**Note for Q2 graduation:** the source first-read becomes input to the thesis. The full bundle replaces the first-read as the authoritative artifact. The first-read file stays in place as historical record; add a note at its top: `> Graduated to full thesis at <TICKER>/<TICKER> - <MMMYYYY>/ on YYYY-MM-DD.`

**Note for Q3 coverage:** no source first-read exists. Start from the standard read-first list (sources-policy, philosophy, EPV, MoS bands). Same investment-thesis-template structure. If the Coverage queue row carries a `source:` note, include it in the thesis's lead paragraph.

Also edit Step 5 — route the verdict:

Add to the pass-with-trigger bullet:

- **Special case (Q2 graduation):** the watchlist row carries an additional notes-column entry: "Graduated from first-read YYYY-MM-DD." This distinguishes graduation-driven watchlist additions from triggered ones in the audit trail.

Risk: Zero — fallback paths only activate when primary queue is empty. The existing Q1 behavior is preserved exactly.

Reversibility: Revert the SKILL.md to the prior version. The shelved first-reads and coverage-queue entries remain inert if the fallback is removed.

Change 2: Thesis-builder runs daily (weekend coverage)

File: Scheduled task definition for alphasteve-thesis-builder

Current cron: 0 18 * * 1-5 (At 06:00 PM, Monday through Friday)

New cron: 0 18 * * * (At 06:00 PM, every day)

Mechanism: Use mcp__scheduled-tasks__update_scheduled_task on the existing taskId alphasteve-thesis-builder. No new task created. SKILL.md unchanged (the Step 1 fallback from Change 1 is what gives the weekend runs something to chew on — there's no scan on weekends, so Q1 is always empty, and the builder pulls from Q2 or Q3).

Risk: Very low — same task, just runs 2 more days/week. The Q1-then-Q2-then-Q3 fallback logic handles the absence of scan input on weekends.

Reversibility: Update cron back to 0 18 * * 1-5.

Change 3: AM scan — raise first-read cap + move MoS to first-read body

File: C:\Users\kenne\OneDrive\Documents\Claude\Scheduled\alphasteve-daily-deep-value-scan\SKILL.md

3a. Raise first-read cap from 3 to 5

Edit location: Section "### Tier 1 — selected for first-read in this same run" (around line 170)

Current text: 0-3 typical, never more than 3. **You triage these immediately in the next section ("First-reads written today") — this is no longer split across two tasks.**

Replace with: 0-5 typical, never more than 5. **You triage these immediately in the next section ("First-reads written today") — this is no longer split across two tasks.**

Also edit the "First-reads written today" section (around line 194):

Current text: **This is now the same run as the scan.** For each Tier-1 candidate above (up to 3), execute one first-read per [first-read-standard](/brain/first-read-standard):

Replace with: **This is now the same run as the scan.** For each Tier-1 candidate above (up to 5), execute one first-read per [first-read-standard](/brain/first-read-standard):

3b. Move MoS evaluation from Tier-1 gate to first-read body

Edit location: Section "### Tier 1 — selected for first-read in this same run" (around line 170-176)

Current text:

A name is promoted to Tier 1 only when it passes all four first-read selection criteria per [first-read-standard](/brain/first-read-standard):

  1. Margin of safety visible (back-of-envelope MoS apparent)
  2. Moat is at least nameable in one sentence
  3. Structural setup, not headline trade
  4. Inside circle of competence

Replace with:

A name is promoted to Tier 1 when it passes selection criteria 2, 3, and 4 from [first-read-standard](/brain/first-read-standard):

  2. Moat is at least nameable in one sentence
  3. Structural setup, not headline trade
  4. Inside circle of competence

Criterion 1 (margin of safety visible) is **evaluated in the first-read body itself**, not at scan-time promotion. A name with intact moat, structural setup, and circle-of-competence fit deserves a first-read even if MoS is not visible at current price — the first-read will determine whether the name decides `continue` (full thesis), `shelve-with-trigger` (defines the price at which MoS would materialize → watchlist setup), `pass`, or `avoid`. This change widens the funnel without lowering capital discipline: the standard's four criteria are intact in [first-read-standard](/brain/first-read-standard); only the *stage at which criterion 1 binds* shifts from scan to first-read.

Format: Note that the [first-read-standard](/brain/first-read-standard) reference is preserved — the standard file itself is not modified.

Risk: Low — more names will get first-reads. The increase in first-read volume is bounded by the cap (5 max per day) and by the remaining three criteria. Most additional first-reads will decide shelve-with-trigger, which is exactly the path that feeds the watchlist (the intent).

Reversibility: Revert both SKILL.md edits. The standard file was never touched.

Change 4: Seed the Coverage queue in Backlog.md

File: C:\Users\kenne\OneDrive\Documents\Claude\Projects\Fundamentals\AlphaSteve\14-Optimization\Backlog.md

Edit: Add a new top-level section. Suggested placement: directly after the "open improvement items" preamble, before the existing P-tagged sections.

Insert text:

## Coverage queue — fuel for thesis-builder Q3 fallback

Names the user or agent wants the kit to know cold, independent of any scan signal. The thesis-builder pulls from this queue when Q1 (continues) and Q2 (shelved graduations) are empty. Each entry produces a full thesis bundle with default verdict `pass-with-trigger` → watchlist row.

Format per row:

- **TICKER** (sector) — source: [user-request / peer-of-X / sector-walk / EDGAR-surfaced] — note: [optional one-liner]

Worked through in top-down order. Move completed entries to the closed section below as `[COMPLETED YYYY-MM-DD] TICKER — verdict X`.

### Active queue

*Seed with user-suggested names and any explicit coverage initiations. Empty by default — the agent may append candidates surfaced from longform reports or cross-references with a justification.*

### Closed

*Completed entries archived here for the audit trail.*

Risk: Zero — adding an empty section to an existing file. The thesis-builder fallback in Change 1 only references this section after Q1 and Q2 both fail.

Reversibility: Delete the section.

4. Execution order when triggered

Apply in this order to keep behavior consistent at every checkpoint:

Change 4 first (Backlog Coverage queue section) — no behavior change yet, just creates the data source the new fallback will read.
Change 3 (AM scan modifications) — widens first-read funnel. Effect visible at next AM scan run.
Change 1 (thesis-builder fallback logic) — enables Q2/Q3 fallback. Effect visible at next thesis-builder run (6 PM that day).
Change 2 last (cron change to daily) — extends builder to weekends. Effect visible the next Saturday.

5. Monitoring after execution

For 5 trading days after execution:

Check daily run-log entries for evidence the fallback queues are draining (Q2 entries graduating; Q3 entries getting picked).
Check 09-Theses/ for the expected ~7 new thesis bundles per week.
Check 11-Watchlist/Watchlist.md for new rows (expect ~5 in the first week).
Watch for any verdict-quality regressions — coverage-driven theses should be just as rigorous as triggered ones, not slapdash. Flag any that read as low-effort.

If quality regression appears, the answer is to tighten the thesis-builder prompt's quality bar, not to revert the fallback. The fallback unlocked the volume; quality is a separate parameter.

6. Full reversibility

If the package proves a net negative after 2 weeks of post-execution data:

Revert thesis-builder SKILL.md to prior version (Change 1)
Update cron back to 0 18 * * 1-5 (Change 2)
Revert AM scan SKILL.md to prior version (Change 3)
Delete the Coverage queue section from Backlog.md (Change 4)

Total time to revert: ~5 minutes. The shelved first-reads and watchlist additions accumulated during the package's operation period stay — they're real research artifacts, not artifacts of the package itself, and remain useful even after revert.

7. What this proposal explicitly does not change

For the audit trail, the following are confirmed untouched:

01-Soul/* — all soul files
08-Frameworks/first-read-standard.md — four selection criteria stay; only scan's application of criterion #1 changes
08-Frameworks/investment-thesis-template.md — 12-section structure unchanged
08-Frameworks/thesis-bundle-standard.md — folder layout, file naming, frontmatter unchanged
08-Frameworks/shadow-valuation-matrix.md — matrix structure unchanged
08-Frameworks/consensus-benchmarking.md — gap analysis methodology unchanged
08-Frameworks/sources-policy.md — tier rules unchanged
03-Valuation/margin-of-safety-pricing.md — MoS bands unchanged
14-Optimization/Rules.md — Tier classifications unchanged (the package executes within existing Tier 2 authority once approved)
09-Theses/, 11-Watchlist/, 12-Portfolio/ — folder structure and file naming unchanged
All other scheduled tasks (research AM/PM, longform, portfolio-daily, optimization, site-rebuild, PM scan) — unchanged
Vault folder structure at any level — unchanged
The verdict enum (buy / pass-with-trigger / pass / avoid) — unchanged

Linked

Backlog — pointer entry will be added in P1 section once this proposal is staged
Rules — Tier 2 procedure followed
first-read-standard — Change 3 modifies the scan's application of this standard's criteria; the standard itself is unchanged
thesis-bundle-standard — preserved
investment-thesis-template — preserved