Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add swan-gtm/gtm-skills --skill reverse-etl-activationgit clone --depth 1 https://github.com/swan-gtm/gtm-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/swan-gtm/gtm-skills/reverse-etl-activation)<a href="https://agentmods.dev/skills/swan-gtm/gtm-skills/reverse-etl-activation"><img src="https://agentmods.dev/badge/skills/swan-gtm/gtm-skills/reverse-etl-activation/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/swan-gtm/gtm-skills/reverse-etl-activation"><img src="https://agentmods.dev/badge/skills/swan-gtm/gtm-skills/reverse-etl-activation.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 25 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
- medium Excessive Agency · line 29 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00194 | $0.01210 |
| Opus 5 | $0.00097 | $0.00605 |
| Sonnet 5 | $0.00039 | $0.00242 |
| Haiku 4.5 | $0.00019 | $0.00121 |
Grade A, and why
reverse-etl-activation scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 13d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 40 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Reverse ETL activation (warehouse to GTM)
The warehouse is the source of truth. This skill moves a computed audience out of it and into the tools where GTM actually happens, without corrupting those tools. The failure mode this skill exists to prevent: a naive "sync the table" job that writes half-matched rows, re-writes unchanged records, blows past API limits, and pushes suppressed or non-consented contacts into a sequence. Do it in this order every time.
Inputs (block and ask if missing)
- Source: warehouse + the exact model/table of the scored/segmented audience (e.g.
analytics.gtm.enterprise_high_intent). Confirm the grain: account, contact, or both. - Destination(s): CRM (Salesforce/HubSpot), ad platform (Meta/Google/LinkedIn), and/or sequencer (Outreach/Salesloft/Swan). Each has its own object model and limits.
- Sync key: the stable identifier used to match a warehouse row to a destination record (domain, email, CRM Id, ad-platform match key).
- Field mapping: which warehouse columns map to which destination fields, and which are read-only in the destination.
Procedure
- Resolve identity BEFORE writing. Map each warehouse account/person to a destination record on the sync key. Apply a match-confidence floor (default
0.90). Rows below the floor are quarantined to a review list, never guessed into a create. Ambiguous many-to-one matches (two warehouse rows, one CRM account) are collapsed by a deterministic rule (most-recent, highest-score), never duplicated. - Compute the change-data-capture (CDC) diff. Compare the resolved audience to the last-synced state. Only real deltas flow: new adds, changed updates (field-level, not whole-row), and intended removals. Unchanged rows never sync (this is what wrecks API budgets and audit logs).
- Map fields with type/format guards. Coerce and validate every mapped field: dates to the destination's format, enums to allowed picklist values, currency/number types, string length caps. A row that fails schema is rejected to an error list with the reason, not force-written.
- Apply suppression + consent gates. Drop anything on the do-not-contact / unsubscribe list, anything failing region/consent (GDPR/CCPA), and anything a channel-specific rule excludes (e.g. no ad-platform push for opted-out contacts). Suppression runs AFTER matching so you can log exactly who was held and why.
- Batch + schedule to the destination's limits. Chunk to each API's batch size and rate limit (default batch
200), backoff on 429s, and schedule cadence per destination (default: CRM near-real-time, ad audiences hourly, sequences gated on an explicit enrollment step, never automatic). - Dry run first. Emit the planned adds / updates / suppressions / errors with counts and a sample of each. Enforce a max-delete guard: if intended removals exceed a share of the audience (default
10%), ABORT and surface it, do not mass-delete on a bad upstream run. Require sign-off unless the run is under an auto-approve threshold. - Live sync, then reconcile. Execute, capture per-row success/failure, and produce a reconciliation report: expected vs. written counts per object, error rows with reasons, suppressed rows. Persist the new sync state so the next run's CDC is correct.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 13d ago First seen · 40 lines · 194 tokens per session scan A ae7af98cd5e8
reverse-etl-activation is a skill published in the GitHub repository swan-gtm/gtm-skills (150 stars, last pushed 2d ago), licensed MIT. It adds 194 tokens to every session and 1,210 once invoked, about $0.0010 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
adoption-leaderboard
Scores a rep's or a team's recent sales calls against a fixed set of five winning-behavior checklists (rapport, upsell signals, renewal risk, customer experience, value realization), then ranks reps lowest-adoption-first so the highest-leverage coaching targets lead. Use when running a team calibration, prepping 1:1…
challenger
Audits a B2B sales call transcript against the Challenger Sale framework (teach, tailor, take control, plus the commercial-teaching pitch arc), or sweeps a CRM export for deals with no teaching insight logged. Use when a rep leans on discovery-question skill but rarely brings a point of view, when prepping a rep for a…
champion-tracker
Tracks champion strength across a deal's call(s) — builds an action ledger, one row per named contact per cited quote, tagging each as either an act of internal selling (looping in a stakeholder, pushing back for the rep in a room the rep wasn't in) or mere positive sentiment (encouraging words, no action), then reads…
end-client-registration
Audits a partner/reseller's end-client deal registration for completeness — partner identity, end-client entity kept distinct from the partner, engagement scope, conflict check against the direct pipeline, confirmed permission to engage the end-client contact, commercial terms, and an expiry window. Use when reviewing…
executive-briefing
Writes the short brief a rep sends upward out of one deal — to their own exec or the buyer's exec sponsor — covering deal snapshot, status, what's at risk, and the ask, each claim cited to a quote or a CSV column/cell. Use when someone who will never read the transcript needs to act on one deal, not a call-by-call…
faint
Audits a B2B sales call transcript against the FAINT qualification framework (Funds, Authority, Interest, Need, Timing), or sweeps a CRM export for deals missing FAINT fields. Use when a prospect hasn't articulated a clear need yet but shows financial capacity and curiosity, or when deciding whether a demand-driven…