Openclix update campaigns
Propose campaign operations (pause/resume/add/delete/update) from measured analytics performance, produce openclix-config.next.json, and apply only after user confirmation. TRIGGER when the user asks to "optimize campaigns", "pause underperforming notifications", "adjust campaigns based on data", or review campaign performance for operational decisions — even if they say "this campaign isn't working, what should I do?". DO NOT trigger for initial campaign design (openclix-design-campaigns), analytics wiring (openclix-analytics), or source template updates (openclix-update).From its SKILL.md
npx -y skills add openclix/openclix --skill openclix-update-campaignsAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
SKILL.md
5.3 KB, as published. Nobody here has run it
OpenClix Update Campaigns
Operate campaigns from measured outcomes, not intuition.
This skill consumes openclix-analytics outputs and generates safe, reviewable campaign updates.
Purpose
- Read campaign performance signals from PA metrics.
- Propose campaign actions per campaign ID.
- Generate
.openclix/campaigns/openclix-config.next.jsonwithout mutating the active config. - Apply changes only after user confirmation.
- Choose update procedure based on detected delivery mode (bundle vs hosted HTTP).
Hard Rules
- Keep default mode as
propose_then_apply. - Never mutate
.openclix/campaigns/openclix-config.jsonbefore user confirmation. - Use campaign-level decisions (
openclix_campaign_id) with conservative sampling. - Follow pause-first policy for deletion:
- no immediate delete from running state
- delete only when already long-paused and repeatedly underperforming
- Inherit config delivery/runtime constraints from
openclix-design-campaigns. - Before writing outputs under
.openclix/**, ensure.openclix/exists in.gitignore.
Inputs
Required:
.openclix/analytics/impact-metrics.json.openclix/campaigns/openclix-config.json
Optional:
.openclix/analytics/campaign-metrics.json.openclix/campaigns/app-profile.json.openclix/campaigns/update-history.json
Outputs
.openclix/campaigns/update-recommendations.json.openclix/campaigns/openclix-config.next.json.openclix/campaigns/update-history.json
Workflow
1) Preflight
- Confirm OpenClix integration exists in the app repository.
- Confirm required input artifacts are present.
- Ensure
.openclix/is ignored by git when.openclix/**outputs are created.
2) Detect delivery mode
Run:
bash skills/openclix-update-campaigns/scripts/detect_delivery_mode.sh --root <target-project-root>
Priority:
- explicit user mode (
--mode) if provided - code-based detection
Returned mode is one of:
bundlehosted_httpdualunknown
If mode is unknown, stop and ask the user to choose bundle or hosted HTTP.
3) Ensure campaign metrics
If .openclix/analytics/campaign-metrics.json is missing:
- Generate a stub file with
status: insufficient_data. - Point to provider query recipes for campaign-level extraction.
- Continue with conservative
no_changedecisions.
4) Evaluate campaign actions
Run:
bash skills/openclix-update-campaigns/scripts/evaluate_campaigns.sh --root <target-project-root>
Decision defaults:
- unit: per campaign
- primary metric: global
d7_retention_delta_pp - guardrails: campaign
open_rate,fail_rate,cancel_rate - minimum sample:
delivered >= 200andopened >= 20
Action rules are documented in references/decision-rules.md.
5) Produce recommendation artifacts
The evaluator must output:
update-recommendations.jsonwith action list, reason codes, and proposed patchesopenclix-config.next.jsonwith draft changes only- updated
update-history.json
6) Confirmation gate
Show the recommendation summary and ask for confirmation. Do not apply to active config until confirmation.
7) Apply by delivery mode (after confirmation)
bundle: sync updated source config to bundled runtime resource path and verify local load +OpenClixCampaignManager.replaceConfig(...)path.hosted_http: publish updated config to user-owned HTTPS host and verify URL health.dual: update both hosted primary and local fallback paths consistently.
8) Validate and hand off
Run platform checks and report:
- selected delivery mode + evidence
- proposed/applied action list
- modified file paths
- verification command results
- any insufficient-data blockers
Automation handoff
For multi-agent retention review loops, run:
bash scripts/retention_ops_automation.sh \
--root <target-project-root> \
--agent all \
--delivery-mode auto \
--dry-run
The helper executes delivery mode detection + evaluator orchestration and writes:
.openclix/automation/run-summary.json.openclix/automation/prompts/openclaw.md.openclix/automation/prompts/claude-code.md.openclix/automation/prompts/codex.md
Failure codes from helper script:
10: prerequisite command or required script missing20: no supported PA provider detected21: OpenClix integration not detected30: required input artifact missing31: delivery mode unresolved (unknown)40: evaluator failed
Resources
references/decision-rules.mdreferences/delivery-mode-detection.mdreferences/campaign-metrics-contract.mdreferences/provider-campaign-query-recipes.md../openclix-design-campaigns/SKILL.md../openclix-analytics/SKILL.md