dsh-sightline
Run the following command in DeepSeek Harness:
dsh plugin install Charlie-Wang-03/dsh-sightline
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install Charlie-Wang-03/dsh-sightline in your DeepSeek Harness terminal to install this plugin; the source is hosted at https://github.com/Charlie-Wang-03/dsh-sightline .
About this plugin
A single repository can carry AGENTS.md, CLAUDE.md, nested instructions, user-global settings, and Claude rules all at once, yet DeepSeek Harness, Codex, and Claude Code each discover those sources in their own way. The result is a silent divergence: the same workspace can present a completely different instruction surface to different agents, and you will never notice unless you actively check.
Sightline turns that hidden gap into one compact three-column comparison. It discovers every relevant instruction source under the current directory, resolves the effective surface for each of the three agents, compares them with deterministic structural rules, and renders agreement, divergence, and unknown states in the DSH Web ToolView with explicit evidence badges. DeepSeek Harness data is observed from live session provenance; Codex and Claude Code data is predicted from their documented discovery semantics. Sightline never presents a prediction as an observed runtime fact.
Sightline is local-first and read-only. It does not modify instruction files, does not send file bodies to any external service, does not lint instruction quality, and does not rewrite or synchronize files. It is built for developers who work with multiple coding agents in parallel and need confidence that their workspace instructions are being loaded the way they expect, especially in teams where agent configuration consistency matters.
Use Cases
- Audit which instruction sources each of DSH, Codex, and Claude Code loads from the same repository
- Review a three-column comparison report in the DSH Web ToolView to verify instruction discovery
- Quickly validate whether agent instruction surfaces change after switching working directories
Best For
- Developers who work with multiple coding agents in parallel
- Engineering teams that need confidence in workspace instruction consistency
- Developers exploring the DSH plugin ecosystem
Related Plugins
A method pack that makes AI coding agents plan against your real baseline, prove completion with fresh evidence, and reduce reworks and unsafe changes.
Turns the DeepSeek Harness session into a captain that builds a durable sub-agent team, splits goals into dependency-aware tasks, and coordinates work via direct messages and a live Web UI.
Gives coding agents design judgment, letting Claude Code, Cursor, and 70+ agents generate and iterate high-quality UI, presentations, and graphics right from the terminal.
Run the Pi ecosystem's plugins on DeepSeek Harness, unmodified, via a compatibility layer that implements Pi's public extension ABI on DSH's native services.