fix(mcp): bound dynamic session and protocol values

[skamenan7]

Summary

Bounds MCP session_id and protocol_version values before writing durable metadata or promoted headers. The protocol-version filter result is also omitted above 256 bytes, rather than turning an optional routing fact into a FilterResultSet::set() error.

• filters/src/agentic/mcp/mod.rs: apply MAX_DYNAMIC_VALUE_LEN at every affected promotion boundary.
• filters/src/agentic/mcp/tests.rs: cover 256-byte retention and 257-byte omission for session IDs and protocol versions.

Related issue

Closes #361

Validation

• Unit tests: cargo test -p praxis-ai-filters -- agentic::mcp
• Integration or functional tests: not applicable, narrow unit-level boundary validation
• make lint (not run)

Checklist

• I reviewed every changed line and can explain the change.
• New capabilities include an example config and functional example test. Not applicable, this fixes existing bound validation.
• User-facing behavior and generated documentation are updated. Not applicable, no config or generated documentation changed.
• Performance-sensitive changes include appropriate benchmark or load-test evidence. Not applicable, this adds existing bound checks only.
• Commits are signed and include a Signed-off-by trailer.

Breaking changes

None.

[praxis-bot]

Review Summary

The length-bound additions for session_id and protocol_version are correct and well-tested. One gap: the adjacent name field in promote_filter_results has the same unbounded set() risk that this PR fixes for protocol_version.

Severity	Count
Critical	0
Large	0
Medium	1

---

Automated review by praxis-bot

[praxis-bot]

Review Summary

The length-bound additions are correct and complete. All dynamic values in write_metadata, promote_mcp_headers, and promote_filter_results now have consistent MAX_DYNAMIC_VALUE_LEN guards. The boundary tests at 256 and 257 bytes are thorough, and the previous finding about the missing name guard in promote_filter_results has been addressed.

Severity	Count
Critical	0
Large	0
Medium	1

---

Automated review by praxis-bot

[praxis-bot]

[Medium] These three assert_* helpers are placed before the // Test Utilities separator. Project conventions require test utility functions to live inside the Test Utilities section, after the separator comment. Move assert_dynamic_values_omitted, assert_dynamic_values_promoted, and assert_oversized_value_is_omitted below the separator alongside the existing assert_invalid_params_rejection, make_default_filter, etc.

[praxis-bot]

No new findings beyond prior review.

All dynamic value promotion paths (write_metadata, promote_mcp_headers, promote_filter_results) now have consistent MAX_DYNAMIC_VALUE_LEN guards. The validate_single_header synthesize path already had its guard pre-PR. Test coverage is thorough: boundary tests at 256/257 bytes for session ID and protocol version, plus oversized method and resource URI tests. Test utilities are properly placed in the Test Utilities section.

---

Automated review by praxis-bot

[praxis-bot]

No new findings beyond prior reviews.

All dynamic value promotion paths (write_metadata, promote_mcp_headers, promote_filter_results) have consistent MAX_DYNAMIC_VALUE_LEN guards. Fixed-value outputs (enum as_str(), "true"/"false") correctly skip guards. The validate_single_header synthesize path already had its guard pre-PR. Test coverage is thorough: boundary tests at 256/257 bytes for session ID and protocol version, plus oversized method and resource URI tests. Test utilities are correctly placed after the Test Utilities separator.

Severity	Count
Critical	0
Large	0
Medium	0

---

Automated review by praxis-bot

[praxis-bot]

No new findings since prior review.

The new commit is a merge of main into the feature branch. The PR-specific files (mod.rs, tests.rs) are unchanged. All dynamic value promotion paths (write_metadata, promote_mcp_headers, promote_filter_results) still have consistent MAX_DYNAMIC_VALUE_LEN guards on every dynamic field (method, name, session_id, protocol_version). Fixed-value outputs (kind, session_present) correctly skip guards. Test coverage remains thorough.

Severity	Count
Critical	0
Large	0
Medium	0

---

Automated review by praxis-bot