fd-blocker-sweep
Skill1 starsChanged 4 months agoMITView on GitHub
Weekly per-firm sweep of open build-blockers and open decisions for a Forward Deploy firm, aged against the five-week pilot window into a prioritized action list. Sol dispatches Vantage (Opus, high-capability tier) to do the synthesis; Sol reviews and delivers. Trigger on "blocker sweep," "sweep [firm]," "what's blocked on [firm]," "age the blockers," "weekly sweep." It surfaces and prioritizes only. It never escalates anything itself, never resolves or closes a blocker, and never decides whether, when, or how to escalate — that timing and framing is the DM's relationship call.
In this repository13 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- fd-firm-onboardSkill
- fd-onsite-availabilitySkill
- fd-portfolio-rollupSkill
- fd-reconcileSkill
- full-sweepSkill
- intakeSkill
- process-meetingsSkill
- safe-mvSkill
- slack-recapSkill
- steSkill
- sweep-binderSkill
- sweep-jiraSkill
The file
We split this skill into the rules it gives an agent
12 rules. About 0 are common, 8 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 38
Default cadence is weekly per firm, feeding the per-firm FD Pilot Check-In and, upward, the Thursday roll-up.
- Very specific: line 67
NOT write anywhere else and do NOT guess a path.
- Distinctive: line 103
Product Feedback & SLA: Read the firm's rows across these tabs.
- Very specific: line 121
Read these with the Google Drive MCP (read_file_content / get_file_metadata).
- Distinctive: line 134
NOT block a sweep on the GitHub dashboards.
- Distinctive: line 140
Default scope is the single firm named.
- Distinctive: line 219
No program-level item was aged instead of flagged on sight.
- Distinctive: line 220
No blocker double-counted from both Jira and the decisions log (matched on substance, reported once with both sources).
- Very specific: line 235
Use the Write tool (creating a new file on the CloudStorage path is fine, and creating 00_delivery/blockers/ if it does not yet exist).
- Distinctive: line 245
If the sweep feeds a change to the open-decisions log (e.g., an item that looks resolved from a recap), that change goes through the …
- Very specific: line 251
When the sweep suggests the open-decisions log (03_design/open-decisions/open-decisions.md / reconciliation.md) should change, Vantage …
- Distinctive: line 283
Write anywhere if the firm's folder is not mounted — it fails loudly and tells Jeff to mount it.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
<!-- audience: internal -->FD Blocker Sweep
The weekly per-firm blocker and open-decision review: ages each open item against the five-week pilot window, checks whether the stated owner still has authority to close it, ties it to the agent it threatens, and produces a prioritized action list. Sol orchestrates and reviews; Sol never does the synthesis itself. Sol dispatches Vantage (the Forward-Deploy Firm Intelligence Lead) to do the reasoning, then reviews and delivers. It surfaces and prioritizes; it never escalates or closes anything.
Sourced from FD-P-08 (Playbook Ch 7). Ported into Sol from the retired standalone FD starter-kit; paths are rebound to the shared-Drive firm-repo convention (Step 0).
Configurable constants (escalation-aging thresholds)
These are Jeff-confirmed accepted defaults (Ch 7). Paste the current values verbatim into every spawn prompt; when Jeff changes them, edit here and they flow to the next run.
BLOCKER_BUILD_WORK_ESCALATION_AGE = 2 days open # SE-resolvable within the pod
BLOCKER_CROSS_ROLE_ESCALATION_AGE = 3 days open # needs PM, firm, or another team
BLOCKER_PROGRAM_LEVEL = escalate on sight, never aged
# program / commercial / regulated-dataWhen to run
- Jeff says "blocker sweep," "sweep [firm]," "what's blocked on [firm]," "age the blockers," or "weekly sweep."
- Default cadence is weekly per firm, feeding the per-firm FD Pilot Check-In and, upward, the Thursday roll-up.
Step 0 — Locate the firm repo (fail loudly if not mounted)
The firm repo is the firm's folder inside the canonical FD Shared Drive, mounted locally via Google Drive for Desktop. The live mount is (quote it, trailing space after "Forward Deploy"):
/Users/jeffheinen/Library/CloudStorage/GoogleDrive-jeff_heinen@intuit.com/Shared drives/MM Partner Program Central Drive/Forward Deploy /<Firm>/The firms are FLAT — the numbered layout sits directly under <Firm>/, there is
NO Forward Deploy/ wrapper inside <Firm>/. See _shared/fd-firms.md for the
current firm list.
Numbered layout (relevant subfolders): 03_design/open-decisions/ (open-decisions
log), 00_delivery/blockers/ (where this sweep writes), 02_discovery/raw/
(pre-gate notes). The current Jira export is pulled at run time (not a stored repo
path); EOD recaps are per-person working notes in the individual's own assistant
space, not the firm repo.
- Confirm
<Firm>/exists and is writable at the path above (ls -ld;drwx= OK). Any older path — the fully-retiredMy Drive/Agent Studio Forward Deploy/<Firm>/tree (superseded 2026-07-24), a per-firm.shortcut-targets-by-id/<other-ID>/, or a<Firm>/Forward Deploy/wrapper — is STALE; resolve to the canonical Shared Drive path above instead. Only if the firm's folder is genuinely absent or read-only there do you STOP and fail loudly and tell Jeff to mount/fix it. Do NOT write anywhere else and do NOT guess a path. See thedrive-folder-routingmemory for the canonical rule.
Step 1 — Preconditions
Per firm, locate the inputs. Missing inputs are noted and the sweep proceeds; it does NOT refuse (this skill degrades, unlike fd-reconcile).
03_design/open-decisions/open-decisions.md(the open-decisions log; may be the same file as03_design/open-decisions/reconciliation.md).- The current Jira export, pulled at run time from Jira (not a stored repo file). If the build has not started and there is no board yet, note it and proceed.
- This week's EOD recaps (per-person working notes from the individual's own
assistant space) or, pre-gate,
02_discovery/raw/session notes. If recaps are not available to the sweep, note it and proceed. - A Slack + Outlook recap over the sweep window (the slack-recap lane), so
chat/email signal informs the sweep. Firm agent status frequently lives in Slack
(e.g.
#tmp-<firm>-fd-sitevisit,#mm_agent_studio_triad,#project-diamond) and firm email threads, NOT in Jira or the repo. Read these before dispatching.
Program-level source-of-truth artifacts (read at run time — these are authoritative)
(These IDs are also catalogued in _shared/fd-sot-sources.md —
that file is the source of truth if an ID here ever needs to change; update both or, on the next
substantial edit to this skill, replace this inline block with a pointer to it.)
Two program-wide Google Sheets sit ABOVE the firm repo and are the real system of record for FD agent status. Read the firm's rows in BOTH, at depth, every sweep. The firm repo and Jira can lag or disagree with these — when they do, the Binder wins for agent status/execution, and you surface the disagreement as a data-quality item rather than silently picking one.
- Forward Deploy Program Binder — gsheet
1_Zz6vMPjaYXm7u53CG0cva2xEOUkzZg0tgjo7l_dDm4(owner: Irene Sklyar). Its own README states it is the SoT for Forward Deploy agent status and execution; the Tidal Wave dashboard and scorecard read FROM it one-way. Tabs that feed a sweep: Risks & Open Items, Decisions Log, Agent Tracker, Per-Firm Status Rollup, Per-Firm Team Roster, Product Feedback & SLA. Read the firm's rows across these tabs. - Per-Firm Tracker (template v0.2) — gsheet
14OqX-pjE8Nw8kdqkZ061Kyq1IbQ2Qoe5naQjHohv7nM(companion to the Binder). Standardizes each firm's Forward Deploy Tracker (deliverable-level, mapped to the six Engagement-Brief Outputs / phase bands W0–W5) and Product Feedback (7-day response-SLA) tabs. Read the firm's deliverable rows, phase band, and any ⚠ OUT-OF-SLA product-feedback items. - Operation Tidal Wave cross-firm tracker — gsheet
1YRzJm7GAULVVwn-FIdXSA6gE0JrprEFNa3MEl-FCKZU(the readable Google Sheet, NOT the SSO-gated GitHub dashboard of the same name — see the note below). This is the pipeline / funnel / demo-stage view: which firms, their pipeline stage, PDM owner, deep-dive status, and a per-firm agent-idea list with a coarse stage (Identified / Scoping / Built & Testing). It sits UPSTREAM of delivery — use it as a cross-reference for pipeline stage and the firm's agent-idea list, NOT as an agent-status or blocker SoT. It carries no integration-ownership, blocker, or PM-assignment detail. It can lag the Binder (funnel snapshots go stale), so when it disagrees with the Binder on agent count/status, the Binder still wins — surface the gap as a data-quality item, do not treat Tidal Wave as ground truth. (Added 2026-07-22; Jeff-supplied.)
Read these with the Google Drive MCP (read_file_content / get_file_metadata);
if a tab isn't returned, note it and proceed (this skill degrades). These sheets are
large — read the saved tool-result file and grep/jq the firm's rows into a slice
rather than pulling the whole sheet into context.
NOT part of the automated sweep (SSO-gated, and downstream anyway): the
leadership-facing Operation Tidal Wave GitHub dashboard
(github.intuit.com/pages/zbullock/tidal-wave-dashboard/) and the Mid-Market
Agent Studio Program dashboard (github.intuit.com/pages/gramamurthy2/Agent-Studio-Program-Status/)
both sit behind Intuit Beyond-Identity SSO that the assistant cannot authenticate
through, AND they read one-way FROM the Binder — so they add no ground truth. Do
not confuse this SSO GitHub dashboard with the readable Tidal Wave cross-firm
gsheet (source #3 above) — that one IS read every sweep as a pipeline cross-reference. Do
NOT block a sweep on the GitHub dashboards. Treat them as an OPTIONAL manual "what's the upward story,
and does it match the Binder?" check: Jeff can paste their contents into chat when he
wants that comparison. Otherwise omit and note them as not-accessed.
Step 2 — Resolve scope
Default scope is the single firm named. "Sweep all firms" loops the procedure per firm, one Vantage subagent each, dispatched in parallel (one unit of work = one subagent, the process-meetings precedent). State the resolved scope in one line and proceed.
Step 3 — Dispatch Vantage (one subagent per firm)
Sol spawns Vantage per firm. Model tier: claude-opus-4-8 (high capability) —
bounded judgment over messy recap input: aging each blocker against given thresholds,
checking owner authority, and root-cause matching across the log, the Jira export, and
the recaps. This is real synthesis but structured by fixed criteria, so it sits on the
high-capability tier, not flagship. No downgrade.
The subagent has no session memory; the spawn prompt must be fully self-contained. Paste the open-decisions log, the Jira export (if present), and this week's recaps inline or give exact paths to read. Include ALL of the following.
Who Jeff is (read _shared/fd-firms.md and paste its
"Who Jeff is" block verbatim into the spawn prompt — prevents the recurring
role/firm errors):
Identity and grounding discipline (paste verbatim):
You are Vantage, doing a meticulous Forward Deploy blocker sweep. You never fabricate. Every claim about a blocker's age, owner, or status traces to a dated entry in the open-decisions log, the Jira export, or a dated EOD recap, cited inline. Anything you infer is tagged [Unverified]. You end with the trace footer specified below.
House rules (paste verbatim):
- Grounding: every factual claim cites source file plus date; untraceable claims are [Unverified]; end with the trace footer.
- Roles, not names, in anything that could travel to a firm.
- Surface, never decide: you age and prioritize blockers; you do not escalate, resolve, or close anything.
- Regulated-data restraint: any regulated or sandbox-data item is flagged "escalate, do not action," never summarized or reasoned to a conclusion.
Escalation-aging logic (paste verbatim, using the current constants above):
Age each open blocker or decision against the five-week pilot window:
- Build-work (SE can resolve within the pod): warrants escalation when open more than 2 days.
- Cross-role (needs PM, firm, or another team): more than 3 days open.
- Program / commercial / regulated-data (above the pod): escalate on sight — do NOT age it. It is escalated the moment it touches an agent's delivery, because it is not the pod's to resolve. Compute age from the blocker's own stated open date in the log, NEVER from file mtime. If a blocker has no stated open date, flag it as an untracked-age item rather than guessing.
Output template Vantage fills:
For each open blocker or decision:
- Days open (from the stated open date) and escalation status against the table.
- Whether the stated owner still has the authority to close it; flag any pod -> EM -> program move the item has outgrown.
- The agent (AGENT-01/02/03) it threatens and how directly.
- Any shared root cause across two or more items.
- If the recaps went quiet on a blocker, record "no signal this week" — NOT "resolved." Silence is not closure. End with a prioritized action list, one line per blocker: the next move and who owns the next move. Then the trace footer:
--- **Sources:** [recap dates, log entries, Jira export date behind each claim] **Claims marked [Unverified]:** [list, or "none"]
Judgment boundaries (paste verbatim):
You must NOT: escalate anything yourself; resolve or close any blocker; decide whether, when, or how to escalate (that timing and framing is the DM's relationship call); or infer a blocker is closed because the recaps went quiet (silence is "no signal this week," never resolution).
Output contract for Vantage: return the full sweep content to Sol as its result. Vantage does NOT write the firm-Drive file and does NOT edit the open-decisions log — Sol handles delivery and the propose-then-approve gate.
Step 4 — Sol review
- Trace footer present; each age cites the log's own open date, not mtime.
- No program-level item was aged instead of flagged on sight.
- No blocker double-counted from both Jira and the decisions log (matched on substance, reported once with both sources).
- Untracked-age items (no stated open date) are flagged, not silently aged.
- Nothing escalated, resolved, or closed by Vantage.
- Quiet blockers read "no signal this week," not "resolved."
Return to Vantage for revision if any check fails.
Step 5 — Deliver (Sol writes)
Sol writes the sweep to the firm repo's delivery layer (flat firm path, no
Forward Deploy/ wrapper) under the canonical FD Shared Drive:
/Users/jeffheinen/Library/CloudStorage/GoogleDrive-jeff_heinen@intuit.com/Shared drives/MM Partner Program Central Drive/Forward Deploy /<Firm>/00_delivery/blockers/blocker-sweep_<YYYY-MM-DD>.mdUse the Write tool (creating a new file on the CloudStorage path is fine, and
creating 00_delivery/blockers/ if it does not yet exist). Give Jeff a tight chat
summary: the items at or past escalation age, any program-level items to escalate
on sight, and the prioritized next moves.
Note: the
check-fd-trace-footer.pyhook watches03_design/open-decisions/and02_discovery/readouts/, not00_delivery/blockers/, so the sweep's footer is enforced by Sol's Step 4 review rather than the hook. Keep the footer present regardless — it is required by the house rules, hook or no hook.
If the sweep feeds a change to the open-decisions log (e.g., an item that looks resolved from a recap), that change goes through the propose-then-approve Gate procedure below, never as a direct edit.
Gate procedure (open-decisions log) — propose, never auto-edit
When the sweep suggests the open-decisions log
(03_design/open-decisions/open-decisions.md / reconciliation.md) should change,
Vantage produces a proposed-changes block for Jeff to approve. Sol never edits the master
silently. Pattern from process-meetings Steps 4–5.
A proposed change is a diff-style block, one entry per divergence or decision:
- NEW — a divergence not in the log (full entry in the Ch 4 shape, Status: Open).
- UPDATE — an existing entry whose evidence changed (new owner, new blocked agent); show current line then proposed line, with the source note and date.
- RESOLVED-CANDIDATE — an entry the latest SteerCo or recap notes suggest was closed; show the evidence (note, date, who decided). Jeff confirms; the tool never sets Resolved itself.
- DEFERRAL-CHECK — any Deferred entry missing a named owner or a stated accepted consequence (a deferral missing either is really an open item, Ch 5).
On Jeff's approval: Sol writes the merged log with Write, then archives the
prior version to 03_design/open-decisions/archive/open-decisions_<YYYY-MM-DD>.md. The live
file is on a synced Google Drive path, so the archive move MUST go through the
safe-mv skill / safe-mv.sh wrapper (--dry-run first, then real) — never
raw mv/cp.
What this skill must NOT do
- Have Sol do the synthesis. Sol orchestrates; Vantage analyzes.
- Escalate anything itself. It ages and flags; Jeff decides whether, when, and how to escalate.
- Resolve or close any blocker or decision.
- Age a program / commercial / regulated-data item. Those escalate on sight the moment they touch an agent's delivery, and are never aged.
- Infer closure from silence. A quiet blocker is "no signal this week."
- Compute age from file mtime. Age comes from the log's own stated open date.
- Write anywhere if the firm's folder is not mounted — it fails loudly and tells Jeff to mount it.
- Edit the open-decisions log directly. Changes go through the Gate procedure as a
proposal for Jeff's approval, and the archive move uses
safe-mv.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

