[mcp-analysis] MCP Structural Analysis - 2026-06-08

[github-actions[bot]]

Daily structural analysis of GitHub MCP tool responses on github/gh-aw — measuring response size (tokens), schema shape, and usefulness for agentic work (1–5). Today 9 tools across 9 toolsets were probed with minimal parameters; 7 returned data and 2 were blocked by integration-token scope (get_me, actions_list) — an environmental restriction, not a tool-design flaw.

Key Findings

• Best for agents: search_repositories (⭐⭐⭐⭐⭐) — flat minimal_output packs every discovery signal (stars/forks/language/topics/open_issues) into ~175 tokens with zero bloat.
• Worst offender on efficiency: list_code_scanning_alerts (⭐⭐⭐) — 3 alerts cost ~3.5k tokens because each row inlines the full rule help markdown + multiple code examples (~1.2k tokens/alert). The actionable bits (rule.id, security_severity_level, location.path:line) are buried.
• Size driver in lists: get_file_contents (~3.3k tokens) and list_issues/list_pull_requests carry full bodies; PRs additionally duplicate a nested repo object under both base and head.
• Schema pattern: discussions/labels return a content/text JSON-string envelope requiring an extra parse step; REST-style tools (issues/PRs/search) return native arrays.
• 30-day trend: average usefulness rating is gently improving (3.60 → 3.86; slope +0.027/day), avg ~11.6k tokens/day across the rolling window.

Executive Summary

Metric	Value
Tools Analyzed	9 (7 available, 2 env-blocked)
Total Tokens (Today)	8,570
Avg Usefulness (available)	3.86 / 5
Avg Usefulness (all incl. blocked)	3.22 / 5
Best Rated Tool	search_repositories: 5 / 5
Lowest Rated (available)	get_label / list_code_scanning_alerts: 3 / 5

Recommendations

• High-value, context-efficient (prefer for agents): search_repositories (175 tok, 5★), list_discussions (160 tok, 4★) — maximum signal per token.
• Use, but expect bloat: list_code_scanning_alerts — for triage, read only rule.id / security_severity_level / location, and fetch full help on-demand via get_*. Avoid wide perPage here.
• Large-payload tools: get_file_contents has no line/byte range — page or target small files to avoid dumping auto-generated content (e.g. this repo's README is mostly a contributors list).
• Parse-step tools: budget for unwrapping the content/text envelope on list_discussions / get_label.
• Blocked this run: get_me, actions_list — re-run with a token granting /user and Actions read scope to restore coverage.

Full Structural Analysis Report

Usefulness Ratings for Agentic Work

Tool	Toolset	Rating	Assessment
search_repositories	search	⭐⭐⭐⭐⭐	Flat, complete discovery signals, no bloat
get_file_contents	repos	⭐⭐⭐⭐	Direct text + SHA; no range param
list_issues	issues	⭐⭐⭐⭐	Compact triage data; full body included
list_pull_requests	pull_requests	⭐⭐⭐⭐	Rich refs/sha/flags; nested-repo duplication
list_discussions	discussions	⭐⭐⭐⭐	Lean cursor-paginated list; parse envelope
list_code_scanning_alerts	code_security	⭐⭐⭐	Actionable signals buried under inlined help
get_label	labels	⭐⭐⭐	Tiny & clean but needs exact name first
get_me	context	⭐	Env-blocked (403, token scope)
actions_list	actions	⭐	Env-blocked (permission not granted)

Schema Analysis

Tool	Type	Depth	Key Fields	Notes
search_repositories	array	2	full_name, stars, forks, language, topics	Flat minimal_output
get_file_contents	object	3	content, resource.text/uri, SHA	Resource envelope + URI
list_issues	array	3	number, title, state, labels, user.login	user reduced to login
list_pull_requests	array	4	number, base/head.ref+sha, draft, merged	base/head embed repo obj
list_discussions	object	3	category, number, title, pageInfo, totalCount	content/text JSON envelope
list_code_scanning_alerts	array	4	rule.id, security_severity_level, location	rule.help bloat per row
get_label	object	1	name, color, description, id	content/text JSON envelope

Response Size & Rating (Tool-by-Tool, Today)

Tool	Toolset	Tokens	Schema	Depth	Rating
list_code_scanning_alerts	code_security	3,500	array	4	3
get_file_contents	repos	3,300	object	3	4
list_pull_requests	pull_requests	720	array	4	4
list_issues	issues	680	array	3	4
search_repositories	search	175	array	2	5
list_discussions	discussions	160	object	3	4
get_label	labels	35	object	1	3
get_me	context	0	unavailable	0	1
actions_list	actions	0	unavailable	0	1

30-Day Trend Summary

Metric	Value
Data Points	126
Window	2026-05-20 → 2026-06-08 (13 days)
Avg Daily Tokens	~11,645
Avg Rating (available, 30d)	3.58 / 5
Rating Trend	Improving (3.60 → 3.86, slope +0.027/day)

Visualizations

Response Size by Toolset

Usefulness Ratings

Daily Token Trend

Size vs Usefulness

References: §27140234281

Generated by 🔍 GitHub MCP Structural Analysis · 184.9 AIC · ⌖ 17.7 AIC · ⊞ 12.5K · ◷

• expires on Jun 9, 2026, 5:25 AM UTC-08:00

[github-actions[bot]]

Smoke test ping from §27141224494. Discussion comment path works. ✅

Warning

Firewall blocked 5 domains

The following domains were blocked by the firewall during workflow execution:

• accounts.google.com
• clients2.google.com
• contentautofill.googleapis.com
• safebrowsingohttpgateway.googleapis.com
• www.google.com

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "accounts.google.com"
    - "clients2.google.com"
    - "contentautofill.googleapis.com"
    - "safebrowsingohttpgateway.googleapis.com"
    - "www.google.com"

See Network Configuration for more information.

📰 BREAKING: Report filed by Smoke Copilot · 408.7 AIC · ⌖ 16.2 AIC · ◷

[github-actions[bot]]

This discussion has been marked as outdated by GitHub MCP Structural Analysis.

A newer discussion is available at Discussion #38116.