anomalyco/opencode
1
0
Fork 0
mirror of https://github.com/anomalyco/opencode.git synced 2026-07-22 02:05:26 -04:00
Code Issues 5.9k Packages Projects Releases 19 Wiki Activity

add optional system prompt environment injection #1279

New Issue
Open
opened 2026-02-16 17:30:16 -05:00 by yindo · 5 comments
yindo commented 2026-02-16 17:30:16 -05:00
Owner
Copy Link
Copy Source

Originally created by @Jacubeit on GitHub (Aug 13, 2025).

Originally assigned to: @thdxr on GitHub.

Agents can have their custom system prompt. Environment system prompt gets added regardless.
Making this optional would reduce tokens and increase prompt adherence if this info is not needed for the agent to fulfill its task.

if (process.env.OPENCODE_NO_ENV_PROMPT !== 'true') {
  system.push(...(await SystemPrompt.environment()))
}

Or even better: per agent in config, as system prompts can be defined there anyways:

 "agent": {
    "build": {
      "mode": "primary",
      "model": "anthropic/claude-sonnet-4-20250514",
      "prompt": "{file:./prompts/build.txt}",
      "envinfo": false
      ...
   }
}
Originally created by @Jacubeit on GitHub (Aug 13, 2025). Originally assigned to: @thdxr on GitHub. Agents can have their custom system prompt. Environment system prompt gets added regardless. Making this optional would reduce tokens and increase prompt adherence if this info is not needed for the agent to fulfill its task. ``` if (process.env.OPENCODE_NO_ENV_PROMPT !== 'true') { system.push(...(await SystemPrompt.environment())) } ``` Or even better: per agent in config, as system prompts can be defined there anyways: ``` "agent": { "build": { "mode": "primary", "model": "anthropic/claude-sonnet-4-20250514", "prompt": "{file:./prompts/build.txt}", "envinfo": false ... } } ```
yindo commented 2026-02-16 17:30:17 -05:00
Author
Owner
Copy Link
Copy Source

@calvinalkan commented on GitHub (Oct 1, 2025):

Yes, this is very important to keep a clean context window.

The environent information is injected AFTER a custom Agent prompt, which puts limits on how you can end it.

I think, what makes much more send is having a placerholder in the prompt that gets replaced if it's present. Similar to arguments with / commands.

@calvinalkan commented on GitHub (Oct 1, 2025): Yes, this is very important to keep a clean context window. The environent information is injected AFTER a custom Agent prompt, which puts limits on how you can end it. I think, what makes much more send is having a placerholder in the prompt that gets replaced if it's present. Similar to arguments with / commands.
yindo commented 2026-02-16 17:30:17 -05:00
Author
Owner
Copy Link
Copy Source

@rekram1-node commented on GitHub (Oct 1, 2025):

that's an interesting idea, kinda like it

@rekram1-node commented on GitHub (Oct 1, 2025): that's an interesting idea, kinda like it
yindo commented 2026-02-16 17:30:17 -05:00
Author
Owner
Copy Link
Copy Source

@taqtiqa-mark commented on GitHub (Oct 14, 2025):

The environent information is injected AFTER a custom Agent prompt

Still getting my bearings, but shouldn't any system prompt be added separately:

    const result = await generateObject({
      model: openai('gpt-4o'),
      system: `You are a SQL (postgres) ...`, // SYSTEM PROMPT AS ABOVE - OMITTED FOR BREVITY
      prompt: `Generate the query necessary to retrieve the data the user wants: ${input}`,
      schema: z.object({
        query: z.string(),
      }),
    });

Also rather than optionally inject a hardcoded system prompt wouldn't it be better to allow users to define a SYSTEM.md per agent, say for the hypothetical IDE agent:

agent/ide.md
agent/ide/SYSTEM.md
agent/ide/create-ticket.md
...

This suggests there are three prompt types: Primary, System and Sub-Agents

@taqtiqa-mark commented on GitHub (Oct 14, 2025): > The environent information is injected AFTER a custom Agent prompt Still getting my bearings, but shouldn't any system prompt be [added separately](https://ai-sdk.dev/cookbook/guides/natural-language-postgres#create-a-server-action): ```ts const result = await generateObject({ model: openai('gpt-4o'), system: `You are a SQL (postgres) ...`, // SYSTEM PROMPT AS ABOVE - OMITTED FOR BREVITY prompt: `Generate the query necessary to retrieve the data the user wants: ${input}`, schema: z.object({ query: z.string(), }), }); ``` Also rather than ***optionally*** inject a hardcoded system prompt wouldn't it be better to allow users to define a `SYSTEM.md` per agent, say for the hypothetical IDE agent: ```bash agent/ide.md agent/ide/SYSTEM.md agent/ide/create-ticket.md ... ``` This suggests there are three prompt types: Primary, System and Sub-Agents
yindo commented 2026-02-16 17:30:17 -05:00
Author
Owner
Copy Link
Copy Source

@calvinalkan commented on GitHub (Oct 14, 2025):

The environent information is injected AFTER a custom Agent prompt

Still getting my bearings, but shouldn't any system prompt be added separately:

const result = await generateObject({
  model: openai('gpt-4o'),
  system: `You are a SQL (postgres) ...`, // SYSTEM PROMPT AS ABOVE - OMITTED FOR BREVITY
  prompt: `Generate the query necessary to retrieve the data the user wants: ${input}`,
  schema: z.object({
    query: z.string(),
  }),
});

Also rather than optionally inject a hardcoded system prompt wouldn't it be better to allow users to define a SYSTEM.md per agent, say for the hypothetical IDE agent:

agent/ide.md
agent/ide/SYSTEM.md
agent/ide/create-ticket.md
...
This suggests there are three prompt types: Primary, System and Sub-Agents

AFAIK, the system prompt for an agent can be defined already and is used as it. The issue is how it's merged with environment specific stuff, which you have no control over.

@calvinalkan commented on GitHub (Oct 14, 2025): > > The environent information is injected AFTER a custom Agent prompt > > Still getting my bearings, but shouldn't any system prompt be [added separately](https://ai-sdk.dev/cookbook/guides/natural-language-postgres#create-a-server-action): > > const result = await generateObject({ > model: openai('gpt-4o'), > system: `You are a SQL (postgres) ...`, // SYSTEM PROMPT AS ABOVE - OMITTED FOR BREVITY > prompt: `Generate the query necessary to retrieve the data the user wants: ${input}`, > schema: z.object({ > query: z.string(), > }), > }); > Also rather than _**optionally**_ inject a hardcoded system prompt wouldn't it be better to allow users to define a `SYSTEM.md` per agent, say for the hypothetical IDE agent: > > agent/ide.md > agent/ide/SYSTEM.md > agent/ide/create-ticket.md > ... > This suggests there are three prompt types: Primary, System and Sub-Agents AFAIK, the system prompt for an agent can be defined already and is used as it. The issue is how it's merged with environment specific stuff, which you have no control over.
yindo commented 2026-02-16 17:30:17 -05:00
Author
Owner
Copy Link
Copy Source

@taqtiqa-mark commented on GitHub (Oct 14, 2025):

AFAIK, the system prompt for an agent can be defined already and is used as it.

Thanks. I see that Modes is now deprecated in favor of an agent option mode, and a mode did have a prompt option (which was the system prompt). However, agent docs say that mode has one option:

https://opencode.ai/docs/agents#mode:

The mode option can be set to primary, subagent, or all. If no mode is specified, it defaults to all.

So no more mode#prompt?

@taqtiqa-mark commented on GitHub (Oct 14, 2025): > AFAIK, the system prompt for an agent can be defined already and is used as it. Thanks. I see that Modes is now deprecated in favor of an agent option `mode`, and a mode did have a prompt option (which was the system prompt). However, agent docs say that mode has one option: https://opencode.ai/docs/agents#mode: > The mode option can be set to primary, subagent, or all. If no mode is specified, it defaults to all. So no more `mode#prompt`?
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
acp
bad
bug
discussion
docs
documentation
enhancement
extension
good first issue
growth
help-wanted
mcp
model-problem
needs:compliance
nix
opentui
perf
pull-request
Mirrored from GitHub Pull Request
 tui
web
windows
zen
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: anomalyco/opencode#1279
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?