Deep RECON — Intel, Familiarization, Critical Evaluation & Proposals
Date: 2026-08-01Branch:
feature/recon-intel-and-navScope: Codebase map, open branches/PRs/commits, refTemplates gaps, navigation SSOT, actionable proposals.
Rule of engagement: expand existing recovery + inventory docs; do not replace them. All write work stays on feature/* until reviewed.
1. Situation report
Canonical pipeline (from
critical-proposal / replit.md):
export_poller.sh, activity_listener.py) are candidates for archive, not dual maintenance.
2. Better navigation than “Entry + HTML only”
_Entry+ReadMe.md and termux-ecosystem-architecture.html are useful but insufficient as the sole pointers. Prefer this SSOT ladder:
Proposal N1: Treat TOOL_INDEX + CONCEPT_INDEX as primary “where do I start?” docs; demote Entry/HTML to secondary.
Proposal N2: Fix REFERENCE_HUB absolute
/data/data/com.termux/... links via archwiz/config.py path roots (align with PR #1 / mistral config work).Proposal N3: Add a short “Navigation SSOT” block at the top of root README (done on this branch) that lists the ladder above — expand-only.
3. Branches — critical evaluation
4. Open PRs — evaluation
PR #1 — Critical Evaluation (docs)
- Keep. Mergeable as documentation.
- Strengths: pipeline truth, config intermediary (not naive relative paths), archive poller/listener, extension table.
- Do not treat poller/listener as first-class in architecture HTML without a “legacy” label.
PR #2 — GHA Rust workflow
- Caution. Monorepo is predominantly Python/Bash; Rust is mainly Harmonizer / select refTemplates.
- Propose: narrow workflow to paths that actually contain Cargo projects; add resource limits suitable for Termux-origin CI expectations.
PR #3 — Session store hygiene (draft)
- Prioritize. Session artifacts in Git is a credential-adjacent risk.
- Requirements before merge: fix pre-commit scanning deleted files to stdout; document rotation of any leaked tokens; history rewrite is separate, reviewed work.
5. refTemplates — CoPilot gaps & nesting debt
5.1 What CoPilot / recovery snapshot listed vs skeleton
Documented categories 01–14 plus loose top-level names. Consolidation commitb104890 removed nested pointers including:
01_Agent_Runtime/deepcode-cli,hermes-agent(later re-skeletoned as metadata)07_Prompt_Context/Interpreted-Context-Methdology(+_fork)13_Third_Party_Refs/assistral15_Reverse_Engineering/entire set: AIStudio2API, AIStudioProxy, AIstudioProxyAPI, gemini-cli-api
5.2 Still uncategorized at tree -L 1 (skeleton branch)
These sit beside 01–14, not inside a numbered category:
5.3 Missing category 15_Reverse_Engineering (removed pointers)
Propose recreate as metadata-only slots (README + SOURCE.txt), same sparse pattern:- AIStudio2API
- AIStudioProxy
- AIstudioProxyAPI
- gemini-cli-api
assistral under 13; deepcode-cli under 01 if still relevant.
5.4 master vs skeleton drift
Proposal R1: Merge skeleton → master as metadata-only (no full clones).
Proposal R2: Add
15_Reverse_Engineering/ placeholders + nest Haven + ICM_fork.Proposal R3: Update root README refTemplates snapshot to mark 15 as intentional category and list uncategorized→nested plan (expand-only).
6. Live codebase comprehension (compressed)
Critical defects already documented (endorse, don’t re-litigate):
- Termux path coupling / broken root symlinks
- Silent
except: passon dispatch (deepcli + multi-ai-cli) - Diverged send_message vs stream_completion payloads
- Hardcoded session UUID fallback
- Session stores tracked in Git (PR #3)
.bakpollution under archwiz
7. Prioritized proposals (“Build the Future Now”)
8. Out of scope / non-goals
- Full recursive submodule update as default (refTemplates = depth-1 sparse metadata).
- Treating
commingle-swarmas production dependency. - History rewrite without explicit, reviewed ops plan (after tip hygiene).
9. Sign-off
This RECON is documentation-only onfeature/recon-intel-and-nav. Implementation of P0–P2 should proceed as separate feature/* branches, one concern each, with PRs against master.