Plugins

Session-level verdict on whether compaction would lose unlanded state; structurally refuses per-block screening.

Extends the composer model selector menu with a third Context window row, preset capacities (64K–2M), custom input, provider-wide sync, and adaptive compaction guard.

Expose a running DeepSeek Harness instance over MCP, so an external agent can list, create and continue that instance's own sessions, delegate a turn and collect its result, answer the questions the DSH agent asks, and set how future approval requests resolve.

Shows what each installed skill costs in the model context and hides a skill from the catalog without deleting it. Lists every skill found on disk with the exact catalog line the model receives (description normalized and truncated at the catalog cap), its body size and token estimate; previews the catalog message byte-for-byte as the agent will receive it, compares that with the copy already published in the current session, and reports skills on disk that never reach the catalog (invalid frontmatter, or shadowed by a higher-priority root). Hiding writes disable-model-invocation into frontmatter and keeps the file, so the skill still works via /name. Also imports skills from ClawHub, Skills.sh and GitHub URLs and uninstalls user-owned ones.

Registers one `context_extract` tool for DeepSeek Harness: it compresses a long raw document (financial report, contract, ledger) into dense structured Markdown plus JSON — key metrics, tables, numbered clauses, entities and cross-check alerts — so a downstream pipeline can work from the compressed form instead of re-reading the raw document. Financial statements, legal contracts and business ledgers are detected and handled differently; balance-sheet and cash tie-outs are run and mismatches are flagged as alerts rather than silently fixed. Read-only; never modifies the source file. Requires Python with openpyxl and python-docx.

Per-workspace gate that hides the tool schemas and system prompt sections of selected plugins from the model in directories that do not need them.

Model-driven tool-result archiving for DeepSeek Harness. Results above a configurable size carry a short inline id (\[#a1b2c3d4]); the model archives the ones it no longer needs into the session's own archives/ folder and the context keeps a small recoverable stub with the file path. Originals stay in the append-only session log, so nothing is lost. Ships a settings panel with per-size-bucket recovery stats, thresholds that apply without restart, and a conditional reminder when results are waiting to be archived. Requires dsh-better-sidebar for the statistics panel.

Adds extra working directories to a workspace, each read-write or read-only, from a folder button next to the composer or the `/add-dir`, `/remove-dir` and `/dirs` commands; the model sees them in its runtime context, and `write` and `edit` calls into read-write directories skip the sandbox escalation prompt by default while shell commands still ask.

One switch that takes an MCP server's tool schemas and its instructions block out of the model context for every agent, live and without a restart, built for the 207 tool schemas REA registers whether or not a task is reverse engineering. A pill in the composer dock, /rea on|off|status|toggle, and GET/POST /rea-toggle/state; the position is remembered across restarts and subagents are covered before their first step.

Five production-grade DSH tools bundled into one plugin (8 tool slots total): git regression finder (find introducer, recent changes, bisect, blame age), stack-trace decoder with local code context, YAML/ENV/properties config validator with typo suggestions, cross-platform shell escape (POSIX/PowerShell/cmd), and MyBatis Plus SQL static auditor (N+1, missing @Transactional, LIKE leading %, SELECT *, missing WHERE). Zero npm dependencies, all offline; concurrency-safe except the bisect tool (dsh-when-broke). Each sub-plugin is also installable standalone (dsh-when-broke, dsh-stackdecoder, dsh-config-guard, dsh-cmd-shield, dsh-mp-auditor).

Notes-backed context compression for DeepSeek Harness: replaces LLM-summarized compaction with a model-written persistent notebook plus recoverable raw history, so the rollover boundary itself costs zero model calls.

Hides other memory plugins' context sections and tools inside a chosen DSH preset, so that session gets a clean context. Off by default; ships a read-only diagnostic tool.

Local code memory with automatic source context, symbol lookup, vector search, and dependency graphs for JavaScript and TypeScript projects.

Runs the upstream Headroom compression toolkit (headroom-ai, Apache-2.0) inside DeepSeek Harness over MCP: headroom_compress, headroom_retrieve and headroom_stats are exposed as mcp__headroom__* tools backed by the upstream server, installed on demand by uvx. Compression is lossy but reversible — originals are fetched back by hash from a local CCR store. Also ships a registrar that adds DeepSeek Harness to Headroom's own MCP registry, and a model-header sanitizer for the upstream proxy.

Semantic context watcher for DSH agents: maintains a journey ledger with flash summarization and MinHash novelty detection, folds milestones, and can inject long-term context with a three-level stall response; injection and the stall-2/3 responses are off by default.

Chunked compaction for oversized sessions: takes over only when the live context exceeds 85% of the model window (or the request exceeds 4 MB), summarizes the history in 1.5 MB chunks and merges them; smaller sessions keep the built-in compaction.

Context observability and control for DeepSeek Harness: cross-session token analytics, session-level budget/prompt/tool controls, and optional agent scaffolds. It reuses the official tokenMeter, compaction, systemPrompt, and tools.restrict seams; monitoring is read-only.

Live context-pressure meter in the composer dock. When usage nears the limit it asks the current model for a handoff brief, starts that brief in a fresh session and switches you there, so long work can continue without hitting the window.

Groups the sessions that hand off to each other into one chain in the DSH sidebar — automatic chain detection from the handoff prompt, per-chain numbering, in-place rename, search, drag and multi-select to re-chain, pinning a chain to the top, archiving chains out of the sidebar into a panel-only archive, a per-session context/steps readout, and configurable handoff thresholds that can trigger an automatic handoff (write the next prompt, open a new session, paste it, switch over).

Purges billion-context snapshots left behind after a session is deleted in DeepSeek Harness, and can also clean up cost-meter ledger rows (off by default).

Indexes session transcripts into the Akasha session layer as turns end, records successful compactions, and adds a per-turn rhythm line to the runtime context. Companions: Hindsight (repo memory), OpenViking (user memory), and the built-in session search (see repo README). Setup: clone the repo, copy core/ aside, then replace <AKASHA_DIR> in cordis.patch.yml with its absolute path (or set the AKASHA_DIR env var); plugins fail loudly while the placeholder remains.

A human-in-the-loop curation workbench inside a DSH session: mark whole turns or single messages to keep verbatim or to summarize, rewrite user, assistant, tool-result and injected-context messages in place without a model call, then compress the marked ranges with model-written summaries or the official /compact. A local JSON backup is written before every write to the session.

Multi-model deliberation pipeline for DeepSeek Harness: independent proposals, blind peer review, program-executed evidence verification, immutable revision and independent acceptance on a fresh context.

Live DeepSeek API balance and this session's token usage in the conversation header, with a detail panel for the topped-up / granted balance split, request count, input, output and cache-hit tokens, and context pressure. The API key is read on every query and is never cached, written to disk or returned to the browser.