[daily secrets] Daily Secrets Analysis — 2026-04-21

[github-actions[bot]]

🔐 Daily Secrets Analysis Report

Date: 2026-04-21
Workflow Files Analyzed: 197
Run: §24749494581

---

📊 Executive Summary

Metric	Value
Workflow files analyzed	197
secrets.* references (total lines)	4,406
github.token references	796
Unique secret types	30
Workflows using secrets	197 / 197 (100%)
Workflows with OTEL observability	54

All 197 compiled workflows use secrets — consistent with the architecture where every agent workflow requires at minimum a GitHub token.

---

🛡️ Security Posture

Control	Status	Coverage
Redaction steps (redact_secrets)	✅ Present	197 / 197 (100%)
Token fallback chain	✅ Present	197 / 197 (100%)
Explicit permissions: blocks	✅ Present	197 / 197 (100%)
Secrets in job outputs: blocks	✅ None	0 (clean)

All three baseline security controls are fully deployed across 100% of workflow files. No secrets are leaked through job outputs — secrets only appear in env: blocks where they are consumed directly by steps.

---

🔑 Token Architecture

The token fallback chain (GH_AW_GITHUB_MCP_SERVER_TOKEN || GH_AW_GITHUB_TOKEN || GITHUB_TOKEN) is used in all 197 workflows, providing a consistent least-privilege escalation path:

1. secrets.GH_AW_GITHUB_MCP_SERVER_TOKEN — dedicated MCP server token (narrowest scope)
2. secrets.GH_AW_GITHUB_TOKEN — general purpose PAT
3. secrets.GITHUB_TOKEN — built-in workflow token (fallback)

---

🤖 Engine Distribution (by Secret Usage)

Engine	Workflows	Key Secret
GitHub Copilot	~129	COPILOT_GITHUB_TOKEN (316 refs)
Anthropic Claude	~56	ANTHROPIC_API_KEY (224 refs)
OpenAI / Codex	~14	OPENAI_API_KEY + CODEX_API_KEY (60 refs each)
Google Gemini	~4	GEMINI_API_KEY (4 refs)

---

🔑 Top 10 Secrets by Usage

Rank	Secret Name	Occurrences	Category
1	GITHUB_TOKEN	2,509	GitHub Auth
2	GH_AW_GITHUB_TOKEN	2,420	GitHub Auth (PAT)
3	GH_AW_GITHUB_MCP_SERVER_TOKEN	1,103	MCP Server Auth
4	COPILOT_GITHUB_TOKEN	316	AI Engine (Copilot)
5	ANTHROPIC_API_KEY	224	AI Engine (Claude)
6	GH_AW_OTEL_ENDPOINT	162	Observability
7	OPENAI_API_KEY	60	AI Engine (OpenAI)
8	CODEX_API_KEY	60	AI Engine (Codex)
9	GH_AW_OTEL_HEADERS	54	Observability
10	GH_AW_CI_TRIGGER_TOKEN	46	CI Automation

📋 Full Secret Inventory (30 unique types)

Secret Name	Occurrences	Category
GITHUB_TOKEN	2,509	GitHub Auth
GH_AW_GITHUB_TOKEN	2,420	GitHub Auth
GH_AW_GITHUB_MCP_SERVER_TOKEN	1,103	MCP Server Auth
COPILOT_GITHUB_TOKEN	316	AI Engine
ANTHROPIC_API_KEY	224	AI Engine
GH_AW_OTEL_ENDPOINT	162	Observability
OPENAI_API_KEY	60	AI Engine
CODEX_API_KEY	60	AI Engine
GH_AW_OTEL_HEADERS	54	Observability
GH_AW_CI_TRIGGER_TOKEN	46	CI Automation
GH_AW_SIDE_REPO_PAT	19	Cross-Repo Access
GH_AW_AGENT_TOKEN	15	Agent Auth
TAVILY_API_KEY	13	Web Search
GH_AW_PROJECT_GITHUB_TOKEN	7	Project Board
NOTION_API_TOKEN	6	Notion Integration
GEMINI_API_KEY	4	AI Engine (Gemini)
BRAVE_API_KEY	4	Web Search
DD_SITE	3	Datadog
DD_APPLICATION_KEY	3	Datadog
DD_API_KEY	3	Datadog
SENTRY_OPENAI_API_KEY	2	Sentry/AI
SENTRY_API_KEY	2	Error Tracking
SENTRY_ACCESS_TOKEN	2	Error Tracking
CONTEXT	2	Custom Context
AZURE_TENANT_ID	2	Azure Auth
AZURE_CLIENT_SECRET	2	Azure Auth
AZURE_CLIENT_ID	2	Azure Auth
SLACK_BOT_TOKEN	1	Notifications
GH_AW_PLUGINS_TOKEN	1	Plugins

🔍 Security Checks Detail

Template Expression Usage

github.event.* expressions appear extensively throughout workflow files (2,477 occurrences outside env blocks). These are used in:

• Job if: conditions (safe — GitHub Actions evaluates these securely)
• concurrency.group expressions (safe — event IDs are not user-controlled secret content)
• env: variable assignments mapping event metadata to environment variables (expected pattern)

This is the standard gh-aw compiled workflow pattern and does not represent template injection risk — event properties like issue.number, comment.id, and pull_request.number are integer IDs, not user-controlled text.

Secrets in Job Outputs

0 secrets found in job output definitions. Secrets are only passed via env: blocks to steps that consume them. This is the correct pattern — secrets are never forwarded as job outputs where they could be read by downstream jobs in unexpected ways.

Observability Secrets (OTEL/Datadog/Sentry)

• 54 workflows use GH_AW_OTEL_ENDPOINT/GH_AW_OTEL_HEADERS for OpenTelemetry tracing
• 1 workflow (mcp-inspector) uses Datadog (DD_API_KEY, DD_APPLICATION_KEY, DD_SITE)
• 2 workflows use Sentry integration (daily-otel-instrumentation-advisor, mcp-inspector)

📖 Reference Documentation

• Specification: scratchpad/secrets-yml.md
• Redaction System: actions/setup/js/redact_secrets.cjs

---

🎯 Key Findings

1. 100% security control coverage — Every workflow has redaction, token cascades, and explicit permissions. No gaps.
2. 30 distinct secret types — Growth from integrations (Azure, Datadog, Sentry, Slack, Notion) alongside core AI engine tokens.
3. GitHub token dominance — GITHUB_TOKEN + GH_AW_GITHUB_TOKEN account for ~57% of all secret references (4,929 / 8,713 total if counting token-cascade lines), reflecting the GitHub-centric architecture.
4. No secrets in outputs — Clean separation between secret consumption (env blocks) and job communication (outputs with step references).
5. Multi-engine secret hygiene — Engine-specific secrets (ANTHROPIC_API_KEY, OPENAI_API_KEY, etc.) are scoped to the specific workflows that need them, not broadly applied.

💡 Recommendations

1. Monitor AZURE_ expansion* — Currently only 2 occurrences; if Azure-based workflows grow, consider a dedicated token cascade pattern similar to the GitHub token chain.
2. Consolidate observability secrets — With 54 workflows using OTEL, consider whether GH_AW_OTEL_ENDPOINT and GH_AW_OTEL_HEADERS should be part of the default token cascade pattern.
3. Audit GH_AW_SIDE_REPO_PAT usage — 19 occurrences of cross-repo PAT usage; verify all are necessary and have minimal scopes.

---

Generated: 2026-04-21T22:24:05Z
Workflow: daily-secrets

Generated by Daily Secrets Analysis Agent · ● 500.6K · ◷

• expires on Apr 24, 2026, 10:27 PM UTC

[github-actions[bot]]

This discussion has been marked as outdated by Daily Secrets Analysis Agent.

A newer discussion is available at Discussion #27911.