feat(packet): alternatives+trap guidance in analysis packet; honest failure_count contract#18
Merged
Merged
Conversation
…nalysis packet Mirror the alternatives guidance the failure and review prompt packets already carry into the analysis packet, in both the Python and Node runtimes, and name a trap so planning surfaces flag a seductive but broken option before committing to one approach. Derived from evaluating the ADHD skill (github.com/UditAkhourii/adhd): the divergence primitive already exists as fanout_start, so the real gap was planning content and recording the choice, not new machinery. Also correct escalation.evidence_source in model-capabilities.json (both mirrored copies) from executed_verification_failure_count to host_supplied_executed_verification_failure_count. Nothing auto-derives that count; requested_failure_count only reads --failure-count and MYTHIFY_FAILURE_COUNT, so the old value overstated an automatic source. The Node analysis packet line is folded onto an existing lines.push so mcp-server/src/workflow-tools.js stays at its 1500 non-blank line ceiling. Verified: python 345/345, node 180/180, interop 12/12, surface and rules manifests, runtime size guard, and generated-file sync all green. Co-Authored-By: Claude Opus 4.8 <[email protected]>
|
Bugbot is not enabled for your account, so this pull request was not reviewed. Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs. |
…rides CI npm audit --audit-level=moderate was red on main and on this branch, independent of the feature change: - fast-uri 3.1.3 (high, GHSA-v2hh-gcrm-f6hx) - @hono/node-server <2.0.5 (moderate, GHSA-frvp-7c67-39w9), transitive via @modelcontextprotocol/sdk The latest SDK (1.29.0, already pinned) still depends on the vulnerable @hono/node-server ^1.19.9, so there is no forward SDK bump. npm audit fix --force would downgrade the SDK to 1.24.3 (breaking). Instead, pin the two transitive packages to patched versions via overrides. Mythify uses only StdioServerTransport, so the hono HTTP path is never exercised. Verified: npm audit --audit-level=moderate reports 0 vulnerabilities, npm ci clean, node 180/180, and the MCP server still lists 41 tools matching protocol/surface-manifest.json. Co-Authored-By: Claude Opus 4.8 <[email protected]>
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
1 participant
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.
What
Implements the actionable outcome of an evaluation of the ADHD skill (github.com/UditAkhourii/adhd), a parallel-divergent-frames-then-critic pattern.
1. Analysis prompt packet now carries hard-to-reverse guidance (both runtimes).
The failure and review packets already ask for
2-3 labeled approaches with tradeoffs. This mirrors that into the analysis packet and adds a trap clause (name the one that looks good but is not and why), so planning surfaces flag a seductive-but-broken option before committing.scripts/mythify_router.pybuild_analysis_prompt_packetmcp-server/src/workflow-tools.jsmirror, folded onto an existinglines.pushso the file stays at its 1500 non-blank line ceilingtests/test_mythify.py2. Honest
escalation.evidence_sourcecontract.Changed
executed_verification_failure_counttohost_supplied_executed_verification_failure_countin both mirroredmodel-capabilities.jsoncopies. Nothing auto-derives that count;requested_failure_countreads only--failure-count/MYTHIFY_FAILURE_COUNT, so the old value overstated an automatic source.Why not more
The divergence primitive already exists (
fanout_startis ADHD Phase 1). The real gap was planning content and a recorded choice, not new machinery. Amythify divergecommand, a.mythify/ideasstate family, ADHD's numeric scoring in state, and anadhd-agentdependency were all considered and rejected; rationale is recorded in.mythifydurable state (decisionadhd-leverage-decisionplus three lessons).Verification
All executed locally, green:
python -m unittest discover -s tests-> 345/345mcp-servernpm test-> 180/180python -m unittest tests.test_interop-> 12/12check_surface_manifest.mjs,check_classification_rules_manifest.mjscheck_runtime_source_size.py(workflow-tools.js held at 1500/1500)build_variants.pyandbuild_registry_docs.mjsproduce no diffAn independent codex gpt-5.6-sol (xhigh) read-only review confirmed the fold is formatter-safe and the
evidence_sourcechange is limited to the two JSON copies.🤖 Generated with Claude Code