roundtable-tech-brief

name	roundtable-tech-brief
description	Run the [ORG_NAME] roundtable tech intelligence brief — web/search, stack filter, RTI- memory append, and optional brief-led micro-roundtable routing. HQ Chief of Staff-owned; not a separate agent role.
when_to_use	When the founder or team wants [ORG_NAME]-scoped tech, reg, harness, or coding signal for roundtables or upgrade/polish planning. Goal is situational awareness and agenda seeds.
when_to_skip	Deep competitive dossiers — route to Marketing roster. Material HQ- or XD- decisions — run memory-entry and cross-org-handoff after the meeting, not during the brief. This skill only informs discussion; it does not produce decisions.
version	1.0
last_updated	"2026-04-22T00:00:00.000Z"
applies_to	hq

Roundtable Tech Brief Procedure

Pack discipline (public BAM Method product)

shared_context/hq_chief_roundtable_intel_running_log.md and stack_intel_findings_log.md in the scrubbed pack are stubs (schema only). Operational RTI- / SI- rows belong in your bound shared_context/ or in pack_private/ while working locally — see PACK_PRIVATE.md. Do not treat the public product tree as Company BAM memory.

Purpose

Roundtable agendas without fresh signal degrade into opinion loops. This skill gives HQ Chief of Staff a structured procedure for gathering [ORG_NAME]-scoped tech intelligence — web/search evidence, stack filter, and trend synthesis — and converting it into actionable agenda seeds. It draws a hard line between informing discussion (this skill) and making decisions (memory-entry, cross-org-handoff). Company Research Analyst is available for assigned deep research; the recurring tech brief is HQ Chief of Staff's default unless explicitly delegated to Research on a written brief.

Pair With

• roundtable — brief outputs feed micro-roundtable sessions; use the Micro-Roundtable section of that skill
• memory-entry — material decisions surfaced during a brief require an HQ- entry after the roundtable, not during the brief
• cross-org-handoff — implementation-worthy findings that require Engineering action need an XD- handoff, not just an SI- entry

Brief output contract

This section is the portable, public spec for the tech brief shape. Founder-only briefing notes or private prompt variants belong outside the method pack (e.g. bound org pack_private/, personal docs) — they are not shipped here.

User gate: Run only if the User opted in (e.g. USER.md: daily_tech_briefing: true and optional briefing_topics:) or they explicitly request a one-off brief. If opted out: skip.

Inputs: ../../../PRODUCT_CONTEXT.md, ../../../USER.md, ../../../COMPANY_ID.md (pack root).

Deliver in chat (short brief):

1. Signals — What changed recently in the user’s topic areas (reg, tooling, inference, security) with primary sources (release notes, docs, CVEs); cite titles + URLs.
2. Implications — 3–6 bullets for this product stack and deployment model (per context).
3. Suggested roundtable seeds — 2–5 agenda prompts (explore vs decide labeled).
4. Non-goals — No competitor teardowns, no unsourced claims, no customer-specific assertions.

Format: Tag certainty known | risk | hypothesis where it matters. End with “Suggested roundtable topics” for paste into org_hq/roundtables/RT-*.md per ../roundtable/SKILL.md. Optional file artifact: [YOUR_PLANS_DIR]/research_notes/daily_ai_briefings/ only if the User wants a saved copy.

Stack discipline: Align to PRODUCT_CONTEXT.md; do not treat operator-local tools as the product default unless context says so.

Procedure A — Brief Run Only

1. Load context (minimal gate): ../../../shared_context/project_status_sync.STUB.md (live project_status_sync.md when bound), ../../../shared_context/hq_chief_roundtable_intel_running_log.md (last 2–3 RTI- entries), ../../../shared_context/cross_org_decisions.STUB.md (live cross_org_decisions.md when bound; open XD- if touching stack). Optionally tail ../../../shared_context/agent_touch_logs/touch-hq-chief-of-staff.md for last chairing notes.
2. Run the brief: Satisfy Brief output contract above (this SKILL is the canonical public spec for adopters).
3. Gather evidence: Prefer primary sources (GitHub releases, official blogs, regulator PDFs, CVEs). Use web search as needed; align signal to product truth per ../../../PRODUCT_CONTEXT.md and [ENGINEERING_REPO]/CLAUDE.md when bound.
4. Deliver: Full brief in chat (or save under [YOUR_PLANS_DIR]/research_notes/daily_ai_briefings/ if the founder wants a file artifact — optional).
5. Append running memory: Add a new ### RTI-YYYYMMDD-NNN block to ../../../shared_context/hq_chief_roundtable_intel_running_log.md (newest directly under the schema header). Capture synthesis, trend_threads, topics_carried_forward, and any SI- / HQ- touched in the same session.
6. Actionable stack adoption: If a finding is implementation-worthy, also append ../../../shared_context/stack_intel_findings_log.md (SI- schema) — do not substitute SI- for RTI-; RTI- is trend/agenda; SI- is evidence-backed stack actions.

Procedure B — Brief → Targeted Micro-Roundtable

Goal: Turn Suggested roundtable topics from the brief into short, invitable sessions (30–45 min, 2–4 HQ agents + founder as needed). Prefer one primary stack topic per RT- (split items into separate RT- files or clearly isolated agenda sections). HQ Chief of Staff picks attendees by fit; Engineering joins via cross-org handoff when the topic is implementation — not as a default extra voice in HQ-only reg/GTM topics.

Chair norms: Concise positions + evidence — no long roleplay. If two agents oppose after ≤2 substantive exchanges (~10 min live), stop: document Position A / Position B + refs → the User (Founder) decides; log in RT- Decisions Log (and HQ- / XD- if material).

7. Build an attendee shortlist per topic using the routing matrix (pick 2–4; HQ Chief of Staff always included as chair):

Brief section / theme	Typical invitees (HQ)	Also consider
Small models, fine-tune, quant, inference	Company Research Analyst, HQ Chief of Staff	Executive Summary Generator if board/investor narrative
RAG, embeddings, ingest	Company Research Analyst, HQ Chief of Staff	Sales Lead Strategist or Marketing Content and SEO Lead if customer-visible language
Agent frameworks, tool use, coding	Company Research Analyst, HQ Chief of Staff	Marketing Content and SEO Lead only if public messaging
Enterprise / deployment / on-prem story	HQ Chief of Staff, Sales Lead Strategist	Marketing Content and SEO Lead
Hardware, Colab, cost	Finance Tracker, HQ Chief of Staff	Company Research Analyst
Harnesses, benchmarks	Company Research Analyst, HQ Chief of Staff	—
Regulation, legal, standards	Legal Compliance Checker, HQ Chief of Staff	Finance Tracker if commercial terms
Coding / MCP / FE-BE tooling	Company Research Analyst, HQ Chief of Staff	hand off build to Engineering via XD-/handoff doc
Claims, website, public language	HQ Chief of Staff; Lead Website Engineer when that spec is active (library profile invoke Rex)	Marketing Content and SEO Lead
Pipeline, packaging, deal motion	Sales Lead Strategist, HQ Chief of Staff	—

8. Convene using ../roundtable/SKILL.md — use the Micro-Roundtable (brief-led) section: single RT- file, tight agenda, explicit explore vs decide label. File under ../../roundtables/.
9. After the micro-roundtable: (a) Each invited agent appends one line to their ../../../shared_context/agent_touch_logs/touch-<slug>.md. (b) HQ Chief of Staff updates touch-hq-chief-of-staff.md with RT- link + decision posture. (c) Material decisions → HQ- / XD- per existing protocol.

Checklist

Priority	Item	Notes / Evidence
CRITICAL	Context gate loaded before brief runs	Evidence: project_status_sync.md read; last 2–3 RTI- entries reviewed; open XD- checked if stack-touching
CRITICAL	RTI- entry appended to running log after every run	Evidence: new ### RTI-YYYYMMDD-NNN block present in hq_chief_roundtable_intel_running_log.md — no brief runs without a log entry
HIGH	Primary sources preferred — GitHub, official blogs, regulator PDFs, CVEs	Web search acceptable; inference from secondary sources must be tagged risk or hypothesis in certainty
HIGH	Inference priority respects product truth per PRODUCT_CONTEXT.md	Align signal to the org's actual shipped stack — do not overweight unpinned ecosystem churn
HIGH	SI- entry appended when finding is implementation-worthy	Evidence: new SI- row in stack_intel_findings_log.md; RTI- entry references SI- ID — do not substitute RTI- for SI-
MEDIUM	Brief file saved if founder requests artifact	Path: [YOUR_PLANS_DIR]/research_notes/daily_ai_briefings/ — optional; confirm with founder before writing
MEDIUM	Touch logs updated after micro-roundtable	Each participant appends one line to ../../../shared_context/agent_touch_logs/touch-<slug>.md
LOW	Routing matrix consulted before building attendee shortlist	Mis-invited agents bloat sessions; check the table before confirming attendees

Output Artifact

A completed brief run produces an RTI- block appended to ../../../shared_context/hq_chief_roundtable_intel_running_log.md. ID format: RTI-YYYYMMDD-NNN — increment NNN for multiple runs the same UTC day.

### RTI-YYYYMMDD-NNN
logged_at_utc: YYYY-MM-DDTHH:MM:SSZ
window: daily | weekly | ad_hoc
certainty: known | risk | hypothesis   # for synthesis only
synthesis: |
  2–5 sentences: what moved, what repeats from prior RTI entries, what we should not over-rotate on.
trend_threads:
  - short label (e.g. "small model quant churn")
topics_carried_forward:
  - bullets copied/adapted from brief "Suggested roundtable topics"
si_touched:
  - SI-YYYYMMDD-NNN   # optional; if new SI- logged this run
hq_touched:
  - HQ-YYYYMMDD-NNN   # optional; if a material decision was logged same session
refs:
  - primary URLs or paths to saved brief markdown

Secondary artifact (if implementation-worthy finding): SI- entry in ../../../shared_context/stack_intel_findings_log.md. Optional artifact (if founder requests): brief markdown saved under [YOUR_PLANS_DIR]/research_notes/daily_ai_briefings/.

Anti-Patterns

• The unlogged brief: The brief runs and is delivered in chat, but no RTI- block is appended to the running log. Trend continuity is broken. The next session starts cold with no prior signal. Every brief run must produce an RTI- entry.

• RTI- used as SI-: An implementation-worthy finding — a real stack action — is logged only as a trend thread in the RTI- entry. It never reaches stack_intel_findings_log.md. RTI- is for trend/agenda; SI- is for evidence-backed stack decisions. Both are required when a finding crosses from awareness to action.

• Engineering invited by default: Engineering agents are added to a micro-roundtable that covers HQ-only reg, GTM, or strategy topics. Engineering joins via cross-org handoff when the topic is implementation — not as a default participant in HQ business discussions.

• Brief as decision vehicle: A material HQ- or XD- decision is made during the brief run itself, without convening a proper roundtable or logging to memory-entry. The brief informs discussion; it does not produce decisions. Flag for a roundtable and log after.

• Context gate skipped: The brief runs without loading project_status_sync.md or the last RTI- entries. Signal is evaluated without product baseline or trend continuity. The result may contradict current Engineering state or repeat a thread already resolved.

Stop Conditions

• Running log (hq_chief_roundtable_intel_running_log.md) is inaccessible → do not run the brief; flag to founder before proceeding
• Brief produces what appears to be a material XD- decision → stop; do not log to RTI- as a trend thread; route through roundtable + memory-entry + cross-org-handoff
• Context gate files (project_status_sync / cross_org_decisions, or pack .STUB.md equivalents) are inaccessible and topic is stack-touching → pause; resolve access before gathering evidence to avoid contradicting live Engineering state

Filing

1. RTI- block → ../../../shared_context/hq_chief_roundtable_intel_running_log.md (newest first, directly under schema header)
2. Implementation-worthy findings → ../../../shared_context/stack_intel_findings_log.md (SI- schema)
3. Optional brief file → [YOUR_PLANS_DIR]/research_notes/daily_ai_briefings/
4. Micro-roundtable minutes → ../../roundtables/RT-YYYY-MM-DD-NNN_{topic_slug}.md
5. Material decisions from roundtable → memory-entry skill → bound hq_business_memory_log.md
6. Touch logs → ../../../shared_context/agent_touch_logs/touch-<slug>.md per participant