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 agentmods add commands/spotify/confidence-ai-plugins/migrate-optimizely-execute-flagsgit clone --depth 1 https://github.com/spotify/confidence-ai-pluginsWhat 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 | $0.00050 | $0.00984 |
| Opus 5 | $0.00025 | $0.00492 |
| Sonnet 5 | $0.00010 | $0.00197 |
| Haiku 4.5 | $0.00005 | $0.00098 |
Grade A, and why
migrate-optimizely-execute-flags 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 2d 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.
What it actually says
Treat this slash command as /migrate-optimizely execute flags.
Read skills/migrate-optimizely/SKILL.md (agent-only; do not narrate).
Starting Phase 1 — Flag execute.
Find .claude/plans/optimizely-flag-migration-*.md (newest if several).
If none, run /migrate-optimizely-plan-flags first — do not create
flags from memory. If Overall is not complete, ask resume the
plan vs execute anyway.
Then follow Execute: How It Works → For flag plans (consent gate, then Flag Setup Sequence). Do not run access IAM writes or code transforms.
Catch-all guarantee: after rules for each flag, if it still has zero
enabled targeting rules, automatically add/enable an everyone
catch-all (default variant). Empty Confidence rules do not serve
everyone. See Automatic everyone catch-all in SKILL.md.
Unsupported operators (must stay flagged): before create or rules
import, re-check for exists / substring / regex. Those rules are
BLOCKED in Confidence — list them, do not import them as-is, and do
not clear BLOCKED without a recorded workaround. See UNSUPPORTED-
OPERATOR GATE and Rules operator audit in SKILL.md.
Progress bars (mandatory — must be visible in chat): every long
write loop must show a live █/░ bar in the chat transcript
(not only inside a collapsed shell panel). Required for flag create,
production waterfall / targeting-rules import, catch-alls, and
resolve verify. For bulk runs: write a script file + progress file,
then every ~15–30s paste the latest bar line into a chat reply.
Do not use a giant inline heredoc whose UI shows
… N lines hidden, and do not treat silent waits as progress. See
Execute progress bar in SKILL.md.
Targeting rules (mandatory next step after flag create): after
flag shells exist, stop and suggest importing planned specific
rules (_rulesets / confidenceRules) as the next step — do not
jump to plan code or declare Phase 1 complete. Use this handoff in
chat:
───── Flag create complete ─────────────────────────────
Next (required): import targeting rules
[1] Start targeting-rules import ← suggested
[2] Pause
Then run the import with its own chat-visible bar:
Execute Flags · targeting rules ████… N/TOTAL flag-id · rule-name
Do not only create flags + everyone catch-alls while skipping
planned rules. Do not fold rules into the create bar. Use the
canonical emitter in Production waterfall / targeting-rules import
(optimizely_execute_rules_progress.txt + chat paste every ~15–30s).
Milestone-only ... created 50 rules or collapsed shell output alone
is a bug.
Validate Phase 1 — resolve all (natural next step after rules):
after targeting-rules import finishes, stop and suggest
resolve-verifying every migrated flag for segment match. This
is how Phase 1 is definitively validated — do not call Phase 1
complete or jump to plan code after rules alone. Use this handoff:
───── Targeting rules import complete ──────────────────
Next (required to validate Phase 1): resolve-verify ALL migrated flags
Goal: every flag returns a segment match (expected variant)
[1] Start resolve-verify all flags ← suggested (validates Phase 1)
[2] Pause
Then run Phase 1 resolve gate with a chat-visible bar
(Execute Flags · resolve verify ████… N/TOTAL). Spot-checking 3–5
flags is not enough. NO_SEGMENT_MATCH / empty assignment = fail.
Write .claude/plans/optimizely-flag-resolve-verify-<date>.json.
Do not say Phase 1 validated while any migrated flag fails verify
unless the operator explicitly skips it with a reason. Only then
suggest plan code.
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.
- 2d ago First seen · 90 lines · 50 tokens per session scan A 443d8fe63a1d
migrate-optimizely-execute-flags is a command published in the GitHub repository spotify/confidence-ai-plugins (8 stars, last pushed today), licensed Apache-2.0. It adds 50 tokens to every session and 984 once invoked, about $0.0003 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 commands, from other repositories
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.
implement
Execute the implementation plan by processing and executing all tasks defined in tasks.md.