--- name: skill-weekly-context-update-33152c5d38b3 description: "**Duration:** 10\u201315 minutes **Cadence:** Weekly (Monday morning recommended) **Output:** Updated context files + a brief change log entry" --- # Skill: Weekly Context Update **Duration:** 10–15 minutes **Cadence:** Weekly (Monday morning recommended) **Output:** Updated context files + a brief change log entry --- ## Quick Start ``` Read skills/weekly-update/SKILL.md and run the weekly context update. ``` That's it. Claude will read the current state of the repo, identify what's stale, draft every section that needs updating, and ask you to fill in the parts it can't know. --- ## Purpose Keep the repo accurate without making maintenance feel like a second job. This skill handles the diff — what changed since last week — rather than asking you to rewrite everything from scratch. The output is a set of proposed edits, not final commits. You review, adjust, approve. Claude drafts; you decide. --- ## What Claude Reads Before producing any output, Claude reads: 1. `CLAUDE.md` — current priorities, active campaigns, team focus 2. `context/signal-library.md` — signal performance log and last-updated date 3. `context/icp-definition.md` — ICP evolution log and last-updated date 4. `context/competitor-radar.md` — win/loss patterns and last-updated date 5. All files in `outputs/campaigns/` — active campaigns and their results tables 6. Any files in `outputs/` dated in the last 14 days 7. `outputs/scoring/` -- scored accounts and their signal dates, for decay recalculation --- ## Step 1: Staleness Check Claude identifies which sections are out of date based on last-updated dates and campaign activity. Flag any of the following: | File / Section | Stale if... | |---------------|-------------| | `CLAUDE.md` → Current priorities | Not updated in the last 7 days | | Signal performance log | A campaign has been live for 14+ days with no results recorded | | Campaign results table | Campaign is live but results rows are empty or older than 7 days | | Competitor radar | Last updated more than 60 days ago | | ICP evolution log | Last updated more than 90 days ago | Print a summary of what's stale before drafting anything. --- ## Step 2: Draft Updates For each stale section, draft a proposed update. Use this format: ``` ### [File] — [Section] Last updated: [date] Status: STALE — [reason] CURRENT: [existing content] PROPOSED: [drafted update based on what Claude can infer from the repo] QUESTIONS FOR YOU: - [anything Claude can't know — requires human input] ``` Work through each stale section in order of impact: ### 2a. CLAUDE.md — Current Priorities Draft a new "Current priorities" block based on: - Which campaigns are currently live (from `outputs/campaigns/`) - What signals are active (from `context/signal-library.md`) - What was in last week's priorities (carry forward anything still relevant) Then ask: - What changed this week that's not reflected in the repo? - Any new accounts or segments becoming a focus? - Any campaigns pausing or wrapping up? ### 2b. Signal Performance Log Pull results from any campaign output files updated in the last 7 days. For each active signal: - Count sends, replies, and meetings booked from the campaign results table - Calculate reply rate and meeting rate - Compare to the existing log entry - Draft an updated row if the numbers changed Flag: if a signal has 30+ sends with no meetings booked, note it explicitly. That's a calibration signal. ### 2c. Active Campaign Results For each campaign in `outputs/campaigns/` with a live results table: - Identify the most recent row - Note if it hasn't been updated in 7+ days - Draft a placeholder row for this week with the date filled in and metrics as `[update]` Ask: what are the current reply, meeting, and pipeline numbers for each live campaign? ### 2d. Competitor Radar (if stale) If `context/competitor-radar.md` hasn't been updated in 60+ days: - Flag it - Ask: any competitive deals won or lost since the last update? Any new objections you're hearing that aren't in the battlecard? Do not draft a competitive update without input — competitive intel requires human knowledge. Just surface the flag and ask the question. ### 2e. ICP Evolution Log (if stale) If the evolution log hasn't been updated in 90+ days: - Draft a log entry template pre-filled with today's date - Ask: has anything changed about who you're targeting? A segment that's underperforming? A new company type showing up in pipeline? ### 2f. Signal Decay Re-scoring If any accounts were scored more than 7 days ago and signals are aging past a decay threshold (30, 60, 90, or 180 days), flag them for re-scoring. **What to check:** - Pull scored accounts from `outputs/scoring/` or CRM export - For each account, check signal ages against the decay table in `context/signal-library.md` - Identify accounts where decay would change the tier assignment (e.g., a signal crossing from 30 to 31 days drops from 100% to 75%) **What to draft:** - List of accounts where the tier would change after decay is applied - For each: current score, decayed score, current tier, new tier, recommended action (downgrade sequence, remove from AE pipeline, move to monitor) **Ask the user:** - Confirm tier changes before applying - Flag any accounts where the decay-driven downgrade should be overridden (e.g., active conversation in progress) **Minimum cadence:** Run this check weekly as part of this skill. For teams with 100+ scored accounts, run a full batch re-score monthly using the ICP Scoring skill. See `workflows/signal-routing.md` for the full decay procedure: when to run the batch, how to handle tier changes, and what not to downgrade silently. --- ## Step 3: Confirm and Apply Present all proposed changes in one response. For each: - Show CURRENT vs. PROPOSED side by side - Mark sections that need human input with `[NEEDS YOUR INPUT]` - After confirmation, apply all approved changes to the relevant files Do not apply changes until the user confirms. Do not invent performance data — if numbers aren't in the repo, ask for them. --- ## Output No separate output file. Changes are applied directly to the relevant context files. Add a one-line entry to `outputs/weekly-log.md` (create it if it doesn't exist): ``` YYYY-MM-DD: Updated [list of files changed]. [One sentence on the most significant change.] ``` This log lets you track how the repo has evolved over time without digging through git history. --- ## If You're Using Automated Data Sync If you have scripts in `sync/` that pull campaign metrics from your outbound tool or CRM, run them before this skill: ```bash python3 sync/sync-campaign-results.py python3 sync/sync-signal-performance.py ``` Then run the skill. Claude will read the freshly synced data and incorporate it into the drafted updates automatically — no manual metric entry required. **Running this automatically with OpenClaw:** [OpenClaw](https://openclaw.ai) is an open-source AI agent that runs locally and can trigger this entire workflow on a schedule — pull metrics via browser automation, update the data files, run this skill, and message you on Slack with the draft for review. See the OpenClaw skill registry for a pre-built GTM repo update skill, or configure your own using the `sync/` scripts as the data layer. --- Source attribution: KarlRaf Original project: https://github.com/KarlRaf/gtm-starter-kit License: MIT License information: https://opensource.org/license/mit Original upstream instructions; downloaded from Mega AI Hub. ## Upstream license notice MIT License Copyright (c) 2026 Gabriel Moreira Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. # Mega AI Hub resource metadata Creator / publisher: KarlRaf Original project: https://github.com/KarlRaf/gtm-starter-kit Attribution basis: Upstream repository explicitly recorded in package mirror path Fields of work: Marketing, Product Management Roles: Growth Marketer, Marketing Analyst, Product Manager Specialties: API Tools: GitHub, Slack Country applicability: Country not specified Classification basis: Source catalog labels Instruction SHA-256: 51200afb66fb5d2737bb204ca588e8e89ac1b4c9267aa293863bab7b3c52d333