[finklen-amazon]

Fixes

Summary

Changes

Several served texts told the model something the server or the AWS APIs do not do. This PR corrects them. It changes text only; request and response handling are unchanged, apart from the wording of one empty-result hint.

• Paging limit. business-support said that omitting max_items fetches every page "(no limit)", and cost-anomaly-monitors left the limit out. Both page through the same time-bounded pager as every other tool, which stops at BCM_MCP_PAGINATION_MAX_SECONDS. Both now splice in PAGINATION_LIMIT_TEXT. ri-performance hardcoded "300-second" in its max_pages line, so with the limit set to 2 the same description said both 2 and 300; it now says "the same time limit". The max_items / max_pages arg docs in these tools, and in cost-anomaly, now point at the limit. The business-support has_more note names the limit as a cause too.

• Metrics default (server instructions). "If metrics is omitted, the tool returns UnblendedCost" holds only for cost-explorer. sp-performance coverage sends SpendCoveredBySavingsPlans, and ri-performance coverage sends no metrics, so the API returns hours only. Both coverage operations reject UnblendedCost with a ValidationException. The sentence now says so.

• Graviton prompt. The Compute Optimizer fallback grouped results by implementation_effort, which Compute Optimizer does not return. Each Compute Optimizer recommendation option carries its own migration_effort (VeryLow, Low, Medium, High) and savings. That scale differs from Cost Optimization Hub's implementation_effort (VeryLow to VeryHigh): for the same instance and the same Graviton target, read-only calls returned Low or Medium from Compute Optimizer and VeryHigh from Cost Optimization Hub. The prompt now names migration_effort, says to use the rank 1 option, and says never to mix the two scales.

• sp-explorer filters. The text said any describe_savings_plans filter value that matches no plan returns an empty list. That holds only for well-formed values. The service rejects these with a ValidationException:

  • a term that is not a whole number of seconds ("1yr", "31536000.0");
  • an upfront or commitment that is not a number ("$4.38");
  • a start or end that does not begin with a millisecond timestamp (a date only, or no milliseconds).

  upfront and commitment compare as numbers ("4.380" matches 4.38). Rate tenancy also takes host; capitalized spellings such as Shared match nothing. The empty-result hint now limits "matched exactly" to text values.

Tests:

• New test_time_bounded_tools_state_the_configured_limit (business-support, cost-anomaly-monitors, ri-performance; description and served arg docs, with the limit overridden).
• New test_metrics_default_is_scoped_to_cost_explorer.
• Extended Graviton prompt and sp-explorer description/hint tests.
• 8 of the new or changed tests fail on main and pass here. The rest are regression guards.

Not changed: cost-comparison rejects filters nested 97 or more levels deep. A probe showed some 97-level shapes succeed in the service, but the exact failing shapes are not pinned down yet, so the limit is left as is. Many other tools' max_items arg docs also say "None fetches every page" without the limit (their tool descriptions do state it). That sweep is left for a follow-up.

User experience

Before:

• A model could read "(no limit)" and treat a time-limited partial result as complete.
• It could pass UnblendedCost to a coverage operation and get a ValidationException.
• On the Compute Optimizer path, it found no implementation_effort to group by, or mixed two different effort scales.
• It could report "no matching plans" for a filter the service had in fact rejected as malformed.

After: the descriptions, arg docs, instructions and prompt match what the server and the APIs do. Every claim was checked against a live call (read-only), and each text claim was shown to fail on main and pass on this branch.

Checklist

If your change doesn't seem to apply, please leave them unchecked.

• I have reviewed the contributing guidelines
• I have performed a self-review of this change
• Changes have been tested
• Changes are documented

Is this a breaking change? N

RFC issue number:

Checklist:

• Migration process documented
• Implement warnings (if it can live side by side)

Acknowledgment

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of the project license.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.79%. Comparing base (69e2666) to head (a6c0cfc).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4994      +/-   ##
==========================================
- Coverage   93.79%   93.79%   -0.01%     
==========================================
  Files        1080     1080              
  Lines       94712    94713       +1     
  Branches    15442    15442              
==========================================
  Hits        88833    88833              
  Misses       3505     3505              
- Partials     2374     2375       +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.