feat(memory-security): prevent secret leakage into agent memory with redaction, validation, and diagnostics#1224
Merged
senamakel merged 4 commits intoMay 5, 2026
Conversation
…h redaction, validation, and diagnostics
Contributor
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (5)
🚧 Files skipped from review as they are similar to previous changes (2)
📝 WalkthroughWalkthroughAdds a new ChangesSecret Detection & Sanitization
Sequence DiagramsequenceDiagram
actor User
participant Tool as MemoryStoreTool
participant Safety as Safety Module
participant Writer as Memory Writer
participant Storage as Persistent Storage
User->>Tool: execute(content, namespace, key, ...)
Tool->>Safety: has_likely_secret(content)?
Safety-->>Tool: true
Tool->>User: error("looks like a secret")
Note over Storage: ✗ not persisted
User->>Tool: execute(safe_content, namespace, key, ...)
Tool->>Safety: has_likely_secret(content)?
Safety-->>Tool: false
Tool->>Writer: request store(...)
Writer->>Safety: sanitize_document_input() / sanitize_json() / sanitize_text()
Safety-->>Writer: Sanitized{ value, report }
Writer->>Storage: persist(sanitized value)
Storage-->>Writer: stored
Writer-->>Tool: success
Tool-->>User: success
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Possibly related PRs
Suggested reviewers
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 3
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
src/openhuman/memory/store/unified/fts5.rs (1)
140-143:⚠️ Potential issue | 🟠 Major | ⚡ Quick winAvoid logging the full
session_idafter insert.This still writes every accepted session identifier verbatim to logs. Even with the new secret detector,
session_idis a high-risk correlation field and should be reduced to a count, hash, or other opaque handle.Suggested fix
tracing::debug!( - "[fts5] inserted episodic entry: session={}, role={}", - entry.session_id, - entry.role + "[fts5] inserted episodic entry: session_chars={} role={}", + entry.session_id.chars().count(), + entry.role );As per coding guidelines,
**/*.rs: “never log secrets or full PII”.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@src/openhuman/memory/store/unified/fts5.rs` around lines 140 - 143, The debug log in fts5.rs is emitting the full session identifier; replace the verbatim session_id usage in the tracing::debug call that references entry.session_id (and still mention entry.role if needed) with an opaque value—e.g., log a truncated/hashed form of entry.session_id or a session counter/handle instead—so that the tracing::debug call no longer writes the full PII; update any code that computes the placeholder (hash/truncate/lookup) next to the existing tracing::debug invocation.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@src/openhuman/memory/safety/mod.rs`:
- Around line 303-323: is_sensitive_key currently only matches exact names and a
few suffixes, so keys like "db_password", "secret_key", or "api_secret" are
missed; update is_sensitive_key to check normalized (lowercased,
alphanumeric-only) keys for contains/ends_with patterns such as "password",
"secret", and "key" as well as the existing "token", "apikey", and
"clientsecret" checks (e.g., use normalized.contains("password") ||
normalized.ends_with("password") || normalized.contains("secret") ||
normalized.ends_with("secret") || normalized.ends_with("key") in addition to
existing matches) so common variants are masked; keep using the same normalized
variable and function name to locate and modify the logic.
- Around line 256-301: sanitize_json_inner currently recurses unbounded; add a
depth guard by changing sanitize_json_inner to accept a depth parameter (e.g.,
fn sanitize_json_inner(value: &Value, depth: usize) -> Sanitized<Value>) and
introduce a constant MAX_SANITIZE_DEPTH; on each recursive call pass depth + 1,
and if depth >= MAX_SANITIZE_DEPTH stop recursing and return a safe replacement
(e.g., Value::String(REDACTED_SECRET.to_string()) or a small sentinel) while
incrementing an appropriate counter on SanitizationReport (e.g.,
truncation/depth_redactions) so callers of sanitize_json_inner (and
sanitize_text if needed) are updated to call the new signature starting with
depth 0.
In `@src/openhuman/tools/impl/memory/store.rs`:
- Around line 95-104: The current safety rejection logs raw namespace and key
values (in the branch calling safety::has_likely_secret(content)), which can
leak secrets; change the branch so it never formats or logs namespace/key
directly—either check namespace and key with safety::has_likely_secret and
reject early, or log only non-sensitive metadata (e.g., content length and
masked identifiers or counts). Update the log::warn! call in the memory_store
safety branch to remove namespace and key from the message (use counts or masked
placeholders), and ensure the function still returns the same ToolResult::error
message without exposing identifiers.
---
Outside diff comments:
In `@src/openhuman/memory/store/unified/fts5.rs`:
- Around line 140-143: The debug log in fts5.rs is emitting the full session
identifier; replace the verbatim session_id usage in the tracing::debug call
that references entry.session_id (and still mention entry.role if needed) with
an opaque value—e.g., log a truncated/hashed form of entry.session_id or a
session counter/handle instead—so that the tracing::debug call no longer writes
the full PII; update any code that computes the placeholder
(hash/truncate/lookup) next to the existing tracing::debug invocation.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: da9491a1-d54e-408f-b38b-d2547f54fc3d
📒 Files selected for processing (9)
src/openhuman/memory/mod.rssrc/openhuman/memory/safety/mod.rssrc/openhuman/memory/store/README.mdsrc/openhuman/memory/store/unified/README.mdsrc/openhuman/memory/store/unified/documents.rssrc/openhuman/memory/store/unified/documents_tests.rssrc/openhuman/memory/store/unified/fts5.rssrc/openhuman/memory/store/unified/kv.rssrc/openhuman/tools/impl/memory/store.rs
AusAgentSmith
pushed a commit
to AusAgentSmith/openhuman
that referenced
this pull request
May 23, 2026
…redaction, validation, and diagnostics (tinyhumansai#1224)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Added centralized memory safety layer to detect and sanitize secret-like values before persistence
Hardened memory persistence boundaries:
memory_docskv_globalkv_namespaceepisodic_logAdded validation in
memory_storetool:Expanded redaction coverage:
Added
[memory:safety]diagnostics:Updated memory-store documentation to reflect new security behavior
Problem
Agent memory may receive sensitive data:
Previous behavior risked:
Protection gaps:
Solution
Introduced shared safety module:
src/openhuman/memory/safety/Capabilities:
Enforcement at Memory Boundaries
Documents
KV Storage
Episodic Memory
Tool-Level Guard
memory_store:Design Tradeoffs
Identifiers
Payloads
Submission Checklist
Tests
docs/TESTING-STRATEGY.mdCoverage
pnpm test:coveragepnpm test:rust.github/workflows/coverage.ymlCoverage Matrix
docs/TEST-COVERAGE-MATRIX.mdFeature IDs
## RelatedNetwork Dependencies
Manual Smoke Checklist
docs/RELEASE-MANUAL-SMOKE.mdLinked Issues
Closes #NNNImpact
Runtime / Platform
Security
Compatibility
Performance
Related
Summary by CodeRabbit
New Features
Tests
Documentation